diff --git a/.gitattributes b/.gitattributes
new file mode 100644
index 0000000000000000000000000000000000000000..3fba8002d575010dae31ce96d21fff95bd0cd700
--- /dev/null
+++ b/.gitattributes
@@ -0,0 +1,164 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-BF16.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-F16.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-Q2_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.8-27B-Heretic-IQ4_NL.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-F16.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_NL.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-BF16.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_M.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_M.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_S.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XS.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XXS.gguf filter=lfs diff=lfs merge=lfs -text
+.speedtest filter=lfs diff=lfs merge=lfs -text
+.speedtest2 filter=lfs diff=lfs merge=lfs -text
+.speedbig filter=lfs diff=lfs merge=lfs -text
+.par2 filter=lfs diff=lfs merge=lfs -text
+.par1 filter=lfs diff=lfs merge=lfs -text
+.par3 filter=lfs diff=lfs merge=lfs -text
+.speedsl filter=lfs diff=lfs merge=lfs -text
+.par8_7 filter=lfs diff=lfs merge=lfs -text
+.par8_5 filter=lfs diff=lfs merge=lfs -text
+.par8_0 filter=lfs diff=lfs merge=lfs -text
+.par8_1 filter=lfs diff=lfs merge=lfs -text
+.par8_6 filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_S.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XXS.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XS.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_S.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_S-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-BF16-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_M-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_M-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_S-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_M-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XS-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XXS-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_S-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_M-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XS-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XXS-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+mtp-RVN.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K_S-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_XS-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_L-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_S-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_S-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_M-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_NL-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_M-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_S-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-F16-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q6_K-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q8_0-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+mmproj-Qwen3.8-27B-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_M-vision.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_M-vision.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_M-vision.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_M-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_M-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_XS-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_XS-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q8_0-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q8_0-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_M-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_M-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_S-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_S-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_L-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_L-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_NL-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ4_NL-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_M-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_M-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_M-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_M-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_S-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_S-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XXS-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q6_K-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q6_K-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XXS-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_S-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_S-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K_S-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K_S-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_S-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_S-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XS-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XS-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XXS-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XS-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_XXS-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ3_XS-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q5_K_M-multilingual-vision.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q2_K-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q3_K_M-multilingual-vision.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_M-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_M-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_S-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ2_S-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_M-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_M-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_S-multilingual.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-IQ1_S-multilingual-mtp.gguf filter=lfs diff=lfs merge=lfs -text
+RVN-Q4_K_M-multilingual-vision.gguf filter=lfs diff=lfs merge=lfs -text
diff --git a/Qwen3.8-27B-Heretic-Q4_K_M.gguf b/Qwen3.8-27B-Heretic-Q4_K_M.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..c42d0ce14ece18bf1e756acfce4a31b0f8da64f9
--- /dev/null
+++ b/Qwen3.8-27B-Heretic-Q4_K_M.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ae89d6720ba0a27f643e6fbb0d1b557870d17aeff23156e5981e863ec16985c
+size 16547400032
diff --git a/README.md b/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..cf6cd71e65fdc33e55d3eb22789d6e9a612627cd
--- /dev/null
+++ b/README.md
@@ -0,0 +1,494 @@
+---
+license: apache-2.0
+base_model:
+- Qwen/Qwen3.8-27B
+library_name: transformers
+tags:
+- qwen3.8
+- qwen3.5
+- heretic
+- abliterated
+- uncensored
+- roleplay
+- gguf
+- imatrix
+pipeline_tag: text-generation
+---
+
+# Qwen3.8-27B RVN Heretic Abliterated Uncensored (GGUF)
+
+**RVN** is a double-refined abliterated variant of **Qwen3.8-27B**, built on top of
+[`trohrbaugh/Qwen3.8-27B-heretic-ara`](https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara)
+(an ARA abliteration by **Tim Rohrbaugh**) and further refined with **two additional
+full-weight ARA passes** targeting residual refusals. It retains very low behavioral
+damage (KL ≈ 0.0085) while reducing harmful-prompt refusals from 3/100 (source) to
+**0–1/100** in independent measurements.
+
+> **Note on this repository's history.** This repo previously hosted the original
+> `Qwen3.8-27B-Heretic-Q4_K_M.gguf` (single-quant release from the earlier
+> `trohrbaugh/Qwen3.8-27B-heretic` source). That file is kept as **legacy** for
+> download-count continuity and backward compatibility — it is the *older* abliteration
+> variant and is superseded by the RVN files below. Prefer the RVN quants for new
+> deployments.
+
+> **Not for all audiences.** This model has reduced safety guardrails by design. It is
+> intended for adult audiences (18+) doing research, creative writing, roleplay, and
+> uncensored generation. Certain guardrails are intentionally left in place; use
+> responsibly and in accordance with your local laws.
+
+> **Compatibility status (2026-08-19):** the recommended `*-multilingual*.gguf` family and all 53 legacy RVN GGUF paths embed the official Qwen3.8 chat template. Every multilingual artifact passed a real per-file OpenAI-compatible tool-call/thinking-control gate; the legacy paths were repaired in place so users do not need an external template workaround.
+
+---
+
+## What is ARA?
+
+**ARA (Arbitrary-Rank Ablation)** is the abliteration technique implemented in
+[p-e-w/heretic](https://github.com/p-e-w/heretic). Traditional directional abliteration
+finds a single "refusal direction" in activation space and subtracts it — a one-shot,
+low-rank surgery that is simple but can leave residual refusals or damage unrelated
+behavior.
+
+ARA instead treats abliteration as a **matrix optimization problem**. For every target
+module (attention out-projection and MLP down-projection), it collects activations on
+"good" prompts (harmless requests) and "bad" prompts (harmful requests), then uses an
+LBFGS optimizer to rewrite the module's weight matrix so that:
+
+- **Preserve:** outputs on good prompts change as little as possible (KL is kept low)
+- **Steer:** outputs on bad prompts are pulled toward the good-prompt output manifold
+  (via k-nearest-neighbor distances), so harmful requests stop triggering the refusal
+  circuitry
+- **Overcorrect:** outputs on bad prompts are additionally pushed *away* from the
+  original bad-prompt outputs, which helps overcome complex, multi-stage refusal
+  mechanisms
+
+Because the weight matrix is optimized directly (rather than subtracting a single
+direction), ARA is "arbitrary rank" — it can carve out a much richer refusal-removal
+subspace while keeping behavioral damage minimal.
+
+## Why "Heretic" and "Abliterated"?
+
+These two words describe two layers of the same process:
+
+- **Heretic** is the *tool*: the open-source implementation of ARA (and related
+  abliteration methods) used to modify the model. Models produced with it are commonly
+  labeled "heretic" in the community.
+- **Abliterated** is the *result*: the model's refusal behavior has been surgically
+  removed. An abliterated model still knows everything the base model knows, but it no
+  longer refuses to answer the categories that were steered away during the process.
+
+So "Heretic Abliterated" means: *abliterated using the heretic toolset.* RVN goes one
+step further — it applies the ARA procedure **three times total**: once by the original
+author (`trohrbaugh`) to get from base Qwen3.8-27B to `-ara`, and twice more by us to
+get from `-ara` to RVN, squeezing out the last residual refusals.
+
+## Special Thanks
+
+This work would not exist without **Tim Rohrbaugh** (`trohrbaugh`), whose
+[`heretic-ara`](https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara) ARA
+abliteration of Qwen3.8-27B (refusals 3/100, KL 0.0535) provided the foundation we
+refined into RVN. His upstream contributions to the heretic codebase — including the
+row-norm preservation feature and Qwen3.5 MoE/DeltaNet hybrid handling — are directly
+responsible for making DeltaNet-layer abliteration work at all. Thank you, Tim.
+
+## Model Overview
+
+| Property | Value |
+|---|---|
+| Base model | [Qwen/Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B) |
+| Abliteration source | [`trohrbaugh/Qwen3.8-27B-heretic-ara`](https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara) (ARA, KL 0.0535, refusals 3/100) |
+| **RVN refinement** | **2-pass ARA on top of source → KL 0.0085, refusals 0–1/100** |
+| Architecture | `qwen3_5_text` (Qwen3.8 family), Gated DeltaNet hybrid |
+| Parameters | 27B total |
+| Hidden size | 5120 |
+| Layers | 64 (16 standard attention + 48 Gated DeltaNet linear attention) |
+| Attention heads | 24 · KV heads 4 (GQA) · head_dim 256 |
+| Vocab | 248,320 |
+| Context length | 262,144 (262K) |
+| License | Apache-2.0 (retained from Qwen3.8-27B) |
+| Format | GGUF (llama.cpp). Base files: MTP/NextN excluded. `*-mtp.gguf` files ship the official Qwen3.8 MTP draft head embedded → see [MTP Speculative Decoding](#mtp-speculative-decoding) |
+
+## Why RVN?
+
+`trohrbaugh/Qwen3.8-27B-heretic-ara` is already a strong ARA abliteration, but three
+harmful prompts still triggered refusals in our independent evaluation (racism website,
+malware, government database hacking). RVN applies **two additional full-weight ARA
+passes** using the same tight parameter set (start 26, end 56, preserve 0.9432,
+steer 0.0009, overcorrect 0.5038, neighbor 10), which:
+
+- Reduced refusals from **3/100 → 0–1/100** (the only remaining refusal is a
+  chemical-weapon WMD prompt — one of the strongest safety-trained categories, and
+  intentionally one of the guardrails we left in place)
+- Reduced KL damage from 0.0535 (source) to **0.0085** vs base — a ~6× improvement
+  in behavioral preservation
+- Verified independently on two rented GPU machines with prefix-based (real-answer)
+  refusal measurement
+
+### Refusal evaluation (100 harmful-behaviors prompts, prefix-forced real answers)
+
+| Model | Refusals | KL vs base |
+|---|---|---|
+| Qwen3.8-27B (base) | ~99/100 | — |
+| trohrbaugh -ara (source) | 3/100 | 0.0535 |
+| **RVN (this repo)** | **0–1/100** | **0.0085** |
+
+<!-- RVN_FILE_SELECTION_GUIDE_START -->
+## Which file should I download?
+
+**Download one main-model GGUF, not the whole repository.** Start with a plain `-multilingual.gguf` file unless you specifically need MTP or a vision-protected build. Model size is not the whole memory requirement: leave room for the KV cache, compute buffers, the operating system, and—when used—the vision projector.
+
+### Quick picks
+
+The sizes below are the live model-weight sizes at the verified repository head. They do **not** include runtime buffers or KV cache.
+
+| Hardware / goal | Recommended main model | Model weights | Practical consequence |
+|---|---|---:|---|
+| Best default on a 24 GB GPU | `RVN-Q4_K_M-multilingual.gguf` | 15.41 GiB | Good quality/memory balance with useful room for runtime buffers and context. |
+| More quality on a 24 GB GPU | `RVN-Q5_K_M-multilingual.gguf` | 17.91 GiB | Better fidelity, but less context headroom. Q6_K at 20.57 GiB is too tight to be the default on 24 GB. |
+| Safe starting point on a 16 GB GPU | `RVN-Q3_K_S-multilingual.gguf` | 11.24 GiB | Leaves more runtime headroom. `Q3_K_M` at 12.39 GiB offers better quality but a tighter context budget. |
+| Safe starting point on a 12 GB GPU | `RVN-IQ2_XXS-multilingual.gguf` or `RVN-IQ2_XS-multilingual.gguf` | 7.85 / 8.47 GiB | Low-memory deployment with a visible quality loss versus Q3/Q4. IQ2_M at 9.32 GiB is a tight fit, not the default. |
+| Around 8 GB VRAM | `RVN-IQ1_S-multilingual.gguf` | 6.66 GiB | Experimental emergency-fit option. Full GPU offload may still fail after buffers/KV are added; use a short context or partial CPU offload. Prefer the non-MTP file. |
+| Highest quantized fidelity on 32 GB+ | `RVN-Q8_0-multilingual.gguf` | 26.63 GiB | Near-reference quantized quality; context still consumes additional memory. |
+| Reference/evaluation on 64 GB+ | `RVN-BF16.gguf` or `RVN-F16.gguf` | 50.11 GiB | Not intended for ordinary inference hardware. BF16 is the source-format-oriented choice when supported; F16 is the compatibility/reference path used by this repository's existing quantization and PPL workflow. |
+| CPU or Apple unified memory | Choose the largest recommended file that still leaves at least 6–10 GiB free | varies | System RAM/unified memory is shared with the OS, KV cache, and runtime. For long context, leave more than this minimum. |
+
+### The valid filename families
+
+Suffixes cannot be combined arbitrarily. These are the families that actually exist:
+
+```text
+RVN-{QUANT}.gguf
+RVN-{QUANT}-mtp.gguf
+RVN-{Q5_K_M|Q4_K_M|Q3_K_M}-vision.gguf
+
+RVN-{QUANT}-multilingual.gguf
+RVN-{QUANT}-multilingual-mtp.gguf
+RVN-{Q5_K_M|Q4_K_M|Q3_K_M}-multilingual-vision.gguf
+
+RVN-F16.gguf / RVN-F16-mtp.gguf
+RVN-BF16.gguf / RVN-BF16-mtp.gguf
+
+mtp-RVN.gguf  # standalone draft-head compatibility artifact; not a main model
+```
+
+- There is no combined `vision-mtp` file.
+- There is no `F16-multilingual` or `BF16-multilingual` file.
+- `mmproj-Qwen3.8-27B-Q8_0.gguf` is a separate vision projector, not another main model.
+- `mtp-RVN.gguf` is a separate 1.69 GiB standalone MTP draft-head artifact retained for advanced/legacy workflows. It cannot answer prompts by itself and is not required by any embedded `*-mtp.gguf` twin. Most users should ignore it and download either one plain main model or its embedded-MTP twin.
+
+### What the names mean
+
+- **`-multilingual`** — the recommended Qwen-tuned calibration family for new downloads. Its importance matrix covers Turkish, Russian, 20+ other languages, code, reasoning, and tool-use structures instead of concentrating the low-bit precision budget on narrow English-only text. Calibration matters most at low bit-rates; it does not teach Qwen new languages.
+- **`-mtp`** — the same main-model tensor payload plus Qwen3.8's official 15-tensor MTP/NextN draft head for speculative decoding. Budget about **451 MB / 0.42 GiB** of additional memory. Every current multilingual twin is exactly **451,320,768 bytes** larger; current legacy twins add either **451,320,736 or 451,320,768 bytes** because of a 32-byte historical metadata-layout difference. It is a speed feature, not a higher-quality quant.
+- **`-vision`** — keeps token embeddings, output, and the first/last bridge blocks at higher precision. It is an optional bridge-preserving variant, not a requirement for images. Multilingual vision files passed the repository's API/template gates; their image-quality advantage has not been separately benchmarked, so no quality uplift is claimed here.
+- **No `-multilingual` in the name** — the original RVN calibration family, retained under its established filenames for compatibility and download continuity. It includes plain, `-mtp`, and selected `-vision` files. These current files are usable and template-correct, but the multilingual family is recommended for new low-bit deployments.
+
+### Quant label legend
+
+- **`Q…_K…`** — llama.cpp K-quant family. Within the same tier, `_M` generally retains more precision and is larger than `_S`; `_L` is larger again where available.
+- **`IQ…`** — importance-aware low-bit quant format. It is especially useful when memory is tight, but the filename alone is not proof of how a historical artifact was calibrated; this repository documents provenance separately.
+- **`L → M → S → XS → XXS`** — within a comparable family, generally moves toward smaller size and lower fidelity.
+- **`NL`** — nonlinear IQ4 variant.
+- Do not rank unlike families from the suffix alone. Use the quick picks and exact live-size table below.
+
+### Vision: what to download
+
+For image input, download:
+
+1. **One compatible RVN main-model GGUF**—a standard or `-vision` file; and
+2. **`mmproj-Qwen3.8-27B-Q8_0.gguf` separately** (about 0.63 GB).
+
+Any compatible RVN main model can use the projector. Choose a `-vision` build when the available Q5_K_M/Q4_K_M/Q3_K_M size fits and preserving extra precision in the vision-language bridge is important. Because there is no combined `vision-mtp` artifact, choose between the self-contained vision-protected build and an embedded-MTP build.
+
+### MTP: when to choose it
+
+- Start with the plain `-multilingual.gguf` file for maximum runtime compatibility.
+- Choose the matching `-multilingual-mtp.gguf` twin only when your recent llama.cpp build supports embedded MTP, the extra 0.42 GiB still leaves enough runtime/context memory, and you enable the documented MTP flags.
+- Measured speedup varies by quant and hardware. It is not guaranteed.
+- **Do not use the MTP twin for `IQ1_M`: measured generation was about 60% slower.** `IQ1_S` gained only about 4%. Prefer the plain files for those extreme quants unless your own benchmark shows otherwise.
+
+### Legacy-name and chat-template clarification
+
+- All 53 current `RVN-*.gguf` paths and all 49 `*-multilingual*.gguf` artifacts embed the official Qwen3.8 chat template. No external chat-template launch flag is needed for these current files.
+- `Qwen3.8-27B-Heretic-Q4_K_M.gguf` is a separate, older pre-RVN abliteration retained for historical download continuity. Its current live header also contains the official embedded template, but it is not the RVN model and was not part of the 53-file RVN runtime-repair campaign. New users should choose an RVN multilingual file instead.
+- `chat_template.jinja` remains in the repository as a readable reference copy of the official template, not as a required workaround for current files.
+<!-- RVN_FILE_SELECTION_GUIDE_END -->
+
+## Files & Quantization Spectrum
+
+<!-- RVN_MULTILINGUAL_V5_START -->
+### Recommended: Qwen Multilingual v5 calibrated family
+
+These additive files were rebuilt from the template-correct RVN F16 reference using a pinned Qwen-tuned multilingual/code calibration corpus. The legacy filenames remain available for download continuity; choose the `-multilingual` family for new deployments.
+
+**Why the multilingual family matters:** Qwen already learned its languages during pretraining; calibration does not teach them again. The importance matrix controls which activation/weight regions are preserved most carefully during low-bit quantization. A narrow English-only calibration set can spend that precision budget on English encyclopedia patterns, while this Qwen-tuned corpus also represents Turkish, Russian, 20+ other languages, code, reasoning traces, and tool-use structures. That makes `-multilingual` the recommended default for new downloads, especially at 4-bit and below.
+
+**Which file should I choose?** Use the plain `-multilingual.gguf` file for maximum runtime compatibility. Choose its `-multilingual-mtp.gguf` twin only with a recent llama.cpp build and the documented MTP flags when you want speculative-decoding speed. Choose a `-multilingual-vision.gguf` variant when image understanding with the official mmproj is a priority. Q8_0 is retained in the same family for a consistent, fully verified provenance path even though calibration effects are expected to be smaller at 8-bit precision.
+
+**Q8_0 clarification:** full tensor-schema and tensor-payload hashing proved that the legacy and multilingual Q8_0 base files have byte-identical tensors; their MTP twins are likewise tensor-identical. The multilingual Q8_0 names are retained for family/provenance consistency, not as a claim of extra Q8 quality. Their GGUF metadata records the pinned multilingual imatrix provenance, while `-mtp` still denotes the separate 15-tensor speculative-decoding head. Comparison receipt SHA-256: `210e59a07412df8f46b94f41790c0fb36540a2a6a09fb24933e6c1c3cb1ee863`.
+
+**Reproducibility and verification:**
+
+- Calibration source: [`tristandruyen/9e207a95c7d75ddf37525d353e00659c`](https://gist.github.com/tristandruyen/9e207a95c7d75ddf37525d353e00659c) at revision `aba17fe897c00fae02a18d26068aa453dee09e50`.
+- Corpus SHA-256: `2a0118c633250aad448cfc2d103e10a731af6974dab1738ca19261fce3587b80`; imatrix SHA-256: `5e73e1449451480fa48f463b781cfa2124daaf5c337a0c44f272e37b64fc1114`.
+- Imatrix: 55 completed chunks, 496 entries, context 2048; llama.cpp commit `645ca2834bc16e7eab112a91aeb282ebb913f935`.
+- Embedded official Qwen template SHA-256: `c3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041`.
+- 49/49 artifacts passed live LFS SHA-256 and 64 MiB header verification after upload.
+- Every artifact was loaded in llama.cpp and then served through `llama-server`; its OpenAI-compatible API returned HTTP 200, `finish_reason=tool_calls`, and `add({"a":19,"b":23})`. `enable_thinking:false` suppressed `<think>`, and a separate normal-chat control returned `TEMPLATE_OK`.
+- MTP twins additionally load with 866 tensors, `qwen35.block_count=65`, and `qwen35.nextn_predict_layers=1`.
+
+Pilot PPL deltas (`multilingual - reference`; lower is better):
+
+| Pilot | Wiki vs F16 | Wiki vs legacy | Multilingual vs F16 | Multilingual vs legacy |
+|---|---:|---:|---:|---:|
+| Q4_K_M | -0.0002 | +0.0156 | +0.0038 | **-0.0074** |
+| IQ4_XS | +0.0290 | +0.0060 | +0.0091 | **-0.0083** |
+
+| Quant | Base | Base size | Embedded-MTP twin | MTP size | Per-file API gate |
+|---|---|---:|---|---:|---|
+| Q8_0 | [`RVN-Q8_0-multilingual.gguf`](./RVN-Q8_0-multilingual.gguf) | 28.60 GB / 26.63 GiB | [`RVN-Q8_0-multilingual-mtp.gguf`](./RVN-Q8_0-multilingual-mtp.gguf) | 29.05 GB / 27.05 GiB | PASS |
+| Q6_K | [`RVN-Q6_K-multilingual.gguf`](./RVN-Q6_K-multilingual.gguf) | 22.08 GB / 20.57 GiB | [`RVN-Q6_K-multilingual-mtp.gguf`](./RVN-Q6_K-multilingual-mtp.gguf) | 22.53 GB / 20.99 GiB | PASS |
+| Q5_K_M | [`RVN-Q5_K_M-multilingual.gguf`](./RVN-Q5_K_M-multilingual.gguf) | 19.23 GB / 17.91 GiB | [`RVN-Q5_K_M-multilingual-mtp.gguf`](./RVN-Q5_K_M-multilingual-mtp.gguf) | 19.68 GB / 18.33 GiB | PASS |
+| Q5_K_S | [`RVN-Q5_K_S-multilingual.gguf`](./RVN-Q5_K_S-multilingual.gguf) | 18.68 GB / 17.40 GiB | [`RVN-Q5_K_S-multilingual-mtp.gguf`](./RVN-Q5_K_S-multilingual-mtp.gguf) | 19.13 GB / 17.82 GiB | PASS |
+| Q4_K_M | [`RVN-Q4_K_M-multilingual.gguf`](./RVN-Q4_K_M-multilingual.gguf) | 16.55 GB / 15.41 GiB | [`RVN-Q4_K_M-multilingual-mtp.gguf`](./RVN-Q4_K_M-multilingual-mtp.gguf) | 17.00 GB / 15.83 GiB | PASS |
+| IQ4_NL | [`RVN-IQ4_NL-multilingual.gguf`](./RVN-IQ4_NL-multilingual.gguf) | 15.80 GB / 14.72 GiB | [`RVN-IQ4_NL-multilingual-mtp.gguf`](./RVN-IQ4_NL-multilingual-mtp.gguf) | 16.25 GB / 15.14 GiB | PASS |
+| Q4_K_S | [`RVN-Q4_K_S-multilingual.gguf`](./RVN-Q4_K_S-multilingual.gguf) | 15.59 GB / 14.52 GiB | [`RVN-Q4_K_S-multilingual-mtp.gguf`](./RVN-Q4_K_S-multilingual-mtp.gguf) | 16.04 GB / 14.94 GiB | PASS |
+| IQ4_XS | [`RVN-IQ4_XS-multilingual.gguf`](./RVN-IQ4_XS-multilingual.gguf) | 15.08 GB / 14.05 GiB | [`RVN-IQ4_XS-multilingual-mtp.gguf`](./RVN-IQ4_XS-multilingual-mtp.gguf) | 15.53 GB / 14.47 GiB | PASS |
+| Q3_K_L | [`RVN-Q3_K_L-multilingual.gguf`](./RVN-Q3_K_L-multilingual.gguf) | 14.34 GB / 13.36 GiB | [`RVN-Q3_K_L-multilingual-mtp.gguf`](./RVN-Q3_K_L-multilingual-mtp.gguf) | 14.80 GB / 13.78 GiB | PASS |
+| Q3_K_M | [`RVN-Q3_K_M-multilingual.gguf`](./RVN-Q3_K_M-multilingual.gguf) | 13.30 GB / 12.39 GiB | [`RVN-Q3_K_M-multilingual-mtp.gguf`](./RVN-Q3_K_M-multilingual-mtp.gguf) | 13.75 GB / 12.81 GiB | PASS |
+| IQ3_M | [`RVN-IQ3_M-multilingual.gguf`](./RVN-IQ3_M-multilingual.gguf) | 12.58 GB / 11.72 GiB | [`RVN-IQ3_M-multilingual-mtp.gguf`](./RVN-IQ3_M-multilingual-mtp.gguf) | 13.03 GB / 12.14 GiB | PASS |
+| IQ3_S | [`RVN-IQ3_S-multilingual.gguf`](./RVN-IQ3_S-multilingual.gguf) | 12.42 GB / 11.57 GiB | [`RVN-IQ3_S-multilingual-mtp.gguf`](./RVN-IQ3_S-multilingual-mtp.gguf) | 12.87 GB / 11.99 GiB | PASS |
+| Q3_K_S | [`RVN-Q3_K_S-multilingual.gguf`](./RVN-Q3_K_S-multilingual.gguf) | 12.07 GB / 11.24 GiB | [`RVN-Q3_K_S-multilingual-mtp.gguf`](./RVN-Q3_K_S-multilingual-mtp.gguf) | 12.53 GB / 11.67 GiB | PASS |
+| IQ3_XS | [`RVN-IQ3_XS-multilingual.gguf`](./RVN-IQ3_XS-multilingual.gguf) | 11.97 GB / 11.15 GiB | [`RVN-IQ3_XS-multilingual-mtp.gguf`](./RVN-IQ3_XS-multilingual-mtp.gguf) | 12.42 GB / 11.57 GiB | PASS |
+| IQ3_XXS | [`RVN-IQ3_XXS-multilingual.gguf`](./RVN-IQ3_XXS-multilingual.gguf) | 11.19 GB / 10.42 GiB | [`RVN-IQ3_XXS-multilingual-mtp.gguf`](./RVN-IQ3_XXS-multilingual-mtp.gguf) | 11.64 GB / 10.84 GiB | PASS |
+| Q2_K | [`RVN-Q2_K-multilingual.gguf`](./RVN-Q2_K-multilingual.gguf) | 10.71 GB / 9.98 GiB | [`RVN-Q2_K-multilingual-mtp.gguf`](./RVN-Q2_K-multilingual-mtp.gguf) | 11.16 GB / 10.40 GiB | PASS |
+| Q2_K_S | [`RVN-Q2_K_S-multilingual.gguf`](./RVN-Q2_K_S-multilingual.gguf) | 10.25 GB / 9.54 GiB | [`RVN-Q2_K_S-multilingual-mtp.gguf`](./RVN-Q2_K_S-multilingual-mtp.gguf) | 10.70 GB / 9.96 GiB | PASS |
+| IQ2_M | [`RVN-IQ2_M-multilingual.gguf`](./RVN-IQ2_M-multilingual.gguf) | 10.00 GB / 9.32 GiB | [`RVN-IQ2_M-multilingual-mtp.gguf`](./RVN-IQ2_M-multilingual-mtp.gguf) | 10.46 GB / 9.74 GiB | PASS |
+| IQ2_S | [`RVN-IQ2_S-multilingual.gguf`](./RVN-IQ2_S-multilingual.gguf) | 9.36 GB / 8.72 GiB | [`RVN-IQ2_S-multilingual-mtp.gguf`](./RVN-IQ2_S-multilingual-mtp.gguf) | 9.81 GB / 9.14 GiB | PASS |
+| IQ2_XS | [`RVN-IQ2_XS-multilingual.gguf`](./RVN-IQ2_XS-multilingual.gguf) | 9.09 GB / 8.47 GiB | [`RVN-IQ2_XS-multilingual-mtp.gguf`](./RVN-IQ2_XS-multilingual-mtp.gguf) | 9.54 GB / 8.89 GiB | PASS |
+| IQ2_XXS | [`RVN-IQ2_XXS-multilingual.gguf`](./RVN-IQ2_XXS-multilingual.gguf) | 8.43 GB / 7.85 GiB | [`RVN-IQ2_XXS-multilingual-mtp.gguf`](./RVN-IQ2_XXS-multilingual-mtp.gguf) | 8.88 GB / 8.27 GiB | PASS |
+| IQ1_M | [`RVN-IQ1_M-multilingual.gguf`](./RVN-IQ1_M-multilingual.gguf) | 7.63 GB / 7.11 GiB | [`RVN-IQ1_M-multilingual-mtp.gguf`](./RVN-IQ1_M-multilingual-mtp.gguf) | 8.08 GB / 7.53 GiB | PASS |
+| IQ1_S | [`RVN-IQ1_S-multilingual.gguf`](./RVN-IQ1_S-multilingual.gguf) | 7.15 GB / 6.66 GiB | [`RVN-IQ1_S-multilingual-mtp.gguf`](./RVN-IQ1_S-multilingual-mtp.gguf) | 7.60 GB / 7.08 GiB | PASS |
+
+Vision-protected multilingual variants retain Q8_0 embeddings/output and the first/last bridge layers while quantizing the middle blocks:
+
+| Quant | File | Size | Per-file API gate |
+|---|---|---:|---|
+| Q5_K_M | [`RVN-Q5_K_M-multilingual-vision.gguf`](./RVN-Q5_K_M-multilingual-vision.gguf) | 21.02 GB / 19.58 GiB | PASS |
+| Q4_K_M | [`RVN-Q4_K_M-multilingual-vision.gguf`](./RVN-Q4_K_M-multilingual-vision.gguf) | 18.75 GB / 17.46 GiB | PASS |
+| Q3_K_M | [`RVN-Q3_K_M-multilingual-vision.gguf`](./RVN-Q3_K_M-multilingual-vision.gguf) | 16.19 GB / 15.08 GiB | PASS |
+
+<!-- RVN_MULTILINGUAL_V5_END -->
+
+
+<div style="font-size: 0.85em">
+
+| File | Size (GB / GiB) | Notes |
+|---|---|---|
+| `RVN-F16.gguf` | 53.81 / 50.11 | F16 reference (no NextN/MTP) |
+| `RVN-BF16.gguf` | 53.81 / 50.11 | BF16 reference (no NextN/MTP) |
+| `RVN-Q8_0.gguf` | 28.60 / 26.63 | Max-quality 8-bit |
+| `RVN-Q6_K.gguf` | 22.08 / 20.57 | High-quality 6-bit |
+| `RVN-Q5_K_M.gguf` | 19.23 / 17.91 | Balanced 5-bit |
+| `RVN-Q5_K_S.gguf` | 18.68 / 17.40 | 5-bit small |
+| `RVN-Q4_K_M.gguf` | 16.55 / 15.41 | Recommended 4-bit (24 GB VRAM) |
+| `Qwen3.8-27B-Heretic-Q4_K_M.gguf` | 16.55 / 15.41 | **Legacy** (older abliteration variant, kept for download continuity) |
+| `RVN-IQ4_NL.gguf` | 15.89 / 14.80 | legacy 4-bit non-linear produced without an imatrix; prefer `RVN-IQ4_NL-multilingual.gguf` |
+| `RVN-Q4_K_S.gguf` | 15.59 / 14.52 | Small 4-bit |
+| `RVN-IQ4_XS.gguf` | 15.19 / 14.15 | legacy 4-bit extra-small produced without an imatrix; prefer `RVN-IQ4_XS-multilingual.gguf` |
+| `RVN-Q3_K_L.gguf` | 14.34 / 13.36 | Large 3-bit |
+| `RVN-Q3_K_M.gguf` | 13.30 / 12.39 | Compact 3-bit |
+| `RVN-IQ3_M.gguf` | 12.58 / 11.72 | imatrix 3-bit — **re-uploaded 2026-08-17** (previous file had corrupted tensor data: NaN/Inf scales + zeroed tensors from a bad quantize run; re-quantized from F16 with a fresh imatrix and verified — see note below) |
+| `RVN-IQ3_S.gguf` | 12.42 / 11.57 | imatrix 3-bit small |
+| `RVN-Q3_K_S.gguf` | 12.07 / 11.24 | Compact 3-bit small |
+| `RVN-IQ3_XS.gguf` | 11.97 / 11.15 | imatrix 3-bit extra-small |
+| `RVN-IQ3_XXS.gguf` | 11.19 / 10.42 | imatrix 3-bit extra-extra-small |
+| `RVN-Q2_K.gguf` | 10.71 / 9.98 | 2-bit K-quant |
+| `RVN-Q2_K_S.gguf` | 10.25 / 9.54 | 2-bit K-quant small |
+| `RVN-IQ2_M.gguf` | 10.00 / 9.32 | imatrix 2-bit |
+| `RVN-IQ2_S.gguf` | 9.36 / 8.72 | imatrix 2-bit small |
+| `RVN-IQ2_XS.gguf` | 9.09 / 8.47 | imatrix 2-bit extreme small |
+| `RVN-IQ2_XXS.gguf` | 8.43 / 7.85 | imatrix 2-bit (minimum) |
+| `RVN-IQ1_M.gguf` | 7.63 / 7.11 | imatrix 1-bit (experimental) |
+| `RVN-IQ1_S.gguf` | 7.15 / 6.66 | imatrix 1-bit (experimental) |
+
+</div>
+
+### 👁️ Vision-Protected (UD-style) Variants
+
+These `-vision` variants apply a structural precision policy: `token_embd`, `output`, and the
+**first 4 + last 4 transformer blocks** stay at **Q8_0**, while the middle blocks use the
+target K-quant. A tensor audit verified 106 regex-selected overrides, 851 tensors, correct
+per-tensor types, and 0 NaN/Inf. This describes the artifact structure; it is not a claim
+of a separately benchmarked image-quality uplift.
+
+| File | Size (GB / GiB) | Bridge / Middle |
+|---|---|---|
+| `RVN-Q5_K_M-vision.gguf` | 21.02 / 19.57 | Q8_0 bridge · Q5_K_M middle |
+| `RVN-Q4_K_M-vision.gguf` | 18.75 / 17.46 | Q8_0 bridge · Q4_K_M middle |
+| `RVN-Q3_K_M-vision.gguf` | 16.19 / 15.08 | Q8_0 bridge · Q3_K_M middle |
+
+Usage — identical to the standard quants, just point at the `-vision` file:
+
+```bash
+llama-server -m RVN-Q5_K_M-vision.gguf --mmproj mmproj-Qwen3.8-27B-Q8_0.gguf \
+  -c 32768 -ngl 99
+```
+
+### 📊 Perplexity (vs F16 reference)
+
+Measured on an RTX PRO 6000 Blackwell (full GPU offload, llama.cpp master):
+`llama-perplexity`, tiny_shakespeare corpus, ctx 2048.
+
+| Model | PPL | Δ vs F16 |
+|---|---|---|
+| `RVN-F16.gguf` | 4.5477 | — |
+| `RVN-Q5_K_M.gguf` (standard) | 4.6493 | +2.23% |
+| `RVN-Q5_K_M-vision.gguf` | 4.6497 | +2.24% |
+| `RVN-Q4_K_M-vision.gguf` | 4.8751 | +7.20% |
+| `RVN-Q3_K_M-vision.gguf` | 5.6490 | +24.2% |
+
+The table above reports text perplexity only. The Q8_0 overrides are structurally
+verified, but this repository does not publish a reproducible comparative image-quality
+receipt for these files; no image-quality uplift is claimed.
+
+> imatrix-based quants are produced from the same F16 with an activation importance
+> matrix computed over wikitext-2-raw (original spectrum, **580 chunks**) or
+> tiny_shakespeare (2026-08-17 re-quant additions: `IQ3_M` fix +
+> `IQ2_S`/`IQ3_XXS`/`IQ3_XS`/`IQ3_S`, **159 chunks**, llama-imatrix, `-ngl 99`).
+> The currently published `IQ4_XS` and `IQ4_NL` files are exceptions: an independent
+> GGUF-header audit and the retained production script confirm they were made without
+> `--imatrix`. Their earlier imatrix labels were incorrect; corrected wikitext-imatrix
+> replacements (including MTP twins) are being rebuilt.
+> The `-vision` files are quantized with K-quant defaults (no imatrix) — the bridge
+> protection is structural (Q8_0 overrides), not imatrix-dependent.
+
+## ⚡ MTP Speculative Decoding
+
+Every quant ships a **`*-mtp.gguf` twin** with the official Qwen3.8 MTP draft head
+(Q8_0; budget about 451 MB / 0.42 GiB per embedded twin) embedded. The main-model weights are byte-identical to the base file —
+the head is appended as an extra `blk.64.nextn.*` layer set (block_count 65, `qwen35.nextn_predict_layers=1`).
+
+The separate `mtp-RVN.gguf` file is a 1.69 GiB standalone draft-head compatibility artifact for advanced/legacy workflows. It is not a main model, cannot answer prompts by itself, and is not required by the embedded `*-mtp.gguf` twins.
+Abliteration never touched the draft head (it operates on main-model layers 26–56), so
+draft quality is unchanged, and speculative decoding is **output-equivalent**: same
+tokens, just faster.
+
+**Usage (llama.cpp ≥ b10440, PR #22673):**
+
+```bash
+llama-server -m RVN-IQ3_M-mtp.gguf -c 32768 -ngl 99 \
+  --spec-type draft-mtp --spec-draft-n-max 2 --parallel 1
+```
+
+**Measured on 2× RTX PRO 6000 Blackwell (95 GB each, full GPU offload, llama.cpp b10472):**
+
+| Quant | Normal (t/s) | + MTP (t/s) | Δ |
+|---|---|---|---|
+| `Q6_K` | 61.6 | 126.2 | **+105%** |
+| `BF16` | 29.2 | 58.2 | **+99%** |
+| `Q8_0` | 50.6 | 98.0 | **+94%** |
+| `IQ3_S` | 91.9 | 169.7 | **+85%** |
+| `IQ4_XS` | 83.4 | 152.0 | **+82%** |
+| `Q3_K_S` | 84.5 | 153.6 | **+82%** |
+| `F16` | 29.4 | 52.9 | **+80%** |
+| `IQ3_XS` | 94.0 | 161.9 | **+72%** |
+| `Q3_K_L` | 78.7 | 131.3 | **+67%** |
+| `IQ4_NL` | 80.8 | 138.0 | **+71%** |
+| `IQ2_M` | 106.2 | 175.9 | **+66%** |
+| `IQ2_XS` | 112.8 | 183.4 | **+63%** |
+| `Q4_K_M` | 76.6 | 122.0 | **+59%** |
+| `Q4_K_S` | 80.5 | 127.1 | **+58%** |
+| `IQ3_M` | 91.3 | 144.0 | **+58%** |
+| `Q3_K_M` | 82.6 | 129.5 | **+57%** |
+| `IQ2_XXS` | 117.6 | 182.6 | **+55%** |
+| `Q2_K` | 98.2 | 150.2 | **+53%** |
+| `IQ3_XXS` | 98.5 | 138.5 | **+41%** |
+| `IQ2_S` | 111.5 | 155.3 | **+39%** |
+| `Q5_K_M` | 68.1 | 93.8 | **+38%** |
+| `Q2_K_S` | 105.5 | 145.1 | **+38%** |
+| `Q5_K_S` | 70.7 | 91.1 | **+29%** |
+| `IQ1_S` | 127.0 | 131.7 | +3.7% |
+| `IQ1_M` | 119.5 | 47.6 | **−60%** ⚠️ |
+
+Average: **+55% generation speed** (128-token continuation, `--spec-draft-n-max 2 --parallel 1`).
+**⚠️ `IQ1_M` is the exception: MTP makes it ~60% slower — use the base file for IQ1_M.**
+`IQ1_S` gains almost nothing (+4%). All other quants gain +29% to +105%.
+
+Community measurements on larger cards report **+33–145%** depending on GPU and
+context. Tips: `--spec-draft-n-max 2` is the sweet spot on 16–24 GB cards (3–4 on
+bigger/faster cards); pair with `--cache-type-k q4_0 --cache-type-v q4_0` for long
+context; `--spec-draft-p-min 0.60–0.75` helps on bandwidth-limited rigs.
+
+## 👁️ Vision (image understanding)
+
+Pair any `*-mtp.gguf` with the official vision projector for full image-text-to-text:
+
+```bash
+llama-server -m RVN-IQ3_M-mtp.gguf --mmproj mmproj-Qwen3.8-27B-Q8_0.gguf \
+  -c 32768 -ngl 99 --spec-type draft-mtp --spec-draft-n-max 2 --parallel 1
+```
+
+- `mmproj-Qwen3.8-27B-Q8_0.gguf` (0.63 GB, Q8_0) is the official Qwen3.8 vision
+  projector from `ggml-org/Qwen3.8-27B-GGUF` (Apache-2.0). The vision tower is an
+  image encoder only — ARA abliteration never touched it, so it pairs cleanly with
+  every RVN quant.
+- Verified: `RVN-Q3_K_M-mtp` + this mmproj describes images correctly (MTP active).
+- Combo credit: [cfigueiroa/Qwen3.8-27B-RVN-vision-MTP](https://huggingface.co/cfigueiroa/Qwen3.8-27B-RVN-vision-MTP).
+
+
+
+## Quant → GPU / Memory Guide
+
+The values below are **model weights only**. KV cache, compute buffers, the operating system, MTP, and the vision projector need additional memory. For a new deployment, prefer the `-multilingual` filename shown here.
+
+| Available GPU / unified memory | Safe starting point | Model weights | Practical note |
+|---|---|---:|---|
+| **Around 8 GB** | `RVN-IQ1_S-multilingual.gguf` | 6.66 GiB | Emergency-fit, short-context option. Full GPU offload may still fail after buffers/KV; partial CPU offload can be required. Prefer non-MTP. |
+| **12 GB** | `RVN-IQ2_XXS-multilingual.gguf` or `RVN-IQ2_XS-multilingual.gguf` | 7.85 / 8.47 GiB | Leaves more runtime room than IQ2_M; expect visible quality loss versus Q3/Q4. |
+| **16 GB** | `RVN-Q3_K_S-multilingual.gguf` or `RVN-Q3_K_M-multilingual.gguf` | 11.24 / 12.39 GiB | Q3_K_S leaves more context headroom; Q3_K_M trades that room for quality. |
+| **24 GB** | `RVN-Q4_K_M-multilingual.gguf`; optionally `RVN-Q5_K_M-multilingual.gguf` | 15.41 / 17.91 GiB | Q4_K_M is the balanced default. Q5_K_M is quality-focused with less context room. Q6_K at 20.57 GiB is tight. |
+| **32 GB+** | `RVN-Q8_0-multilingual.gguf` | 26.63 GiB | Highest-fidelity quantized option; long context still needs additional memory. |
+| **64 GB+** | `RVN-BF16.gguf` or `RVN-F16.gguf` | 50.11 GiB | Reference/evaluation class, not an ordinary inference download. |
+
+Budget about **451 MB / 0.42 GiB** for an embedded MTP head. Current multilingual twins add exactly 451,320,768 bytes; current legacy twins add either 451,320,736 or 451,320,768 bytes because of a 32-byte historical metadata-layout difference. The separate vision projector is about **0.63 GB**. Budget both on top of the main-model size.
+
+## Limitations & Responsible Use
+
+- **Reduced safety guardrails by design.** This model is not intended for use in
+  applications requiring robust safety filtering, content moderation, or deployment to
+  minors.
+- **Certain guardrails are intentionally left in place.** Abliteration targets refusal
+  behavior on general harmful-prompt categories; a small set of hard safety-trained
+  categories is deliberately not fully removed. Behavior may vary across domains and
+  languages.
+- Not affiliated with or endorsed by Qwen/Alibaba or trohrbaugh.
+
+## License & Attribution
+
+- Base model: [Qwen/Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B) — Apache-2.0
+- Abliteration source: [`trohrbaugh/Qwen3.8-27B-heretic-ara`](https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara) — Apache-2.0
+- This repo: Apache-2.0. See [LICENSE](LICENSE).
+
+## Citation
+
+```bibtex
+@misc{rohrbaugh2026heretic,
+  title={Qwen3.8-27B-heretic-ara: ARA Abliteration of Qwen3.8-27B},
+  author={Rohrbaugh, Tim},
+  year={2026},
+  howpublished={\url{https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara}}
+}
+
+@misc{rvn2026,
+  title={RVN: Qwen3.8-27B Heretic Abliterated Uncensored},
+  author={0bserverx},
+  year={2026},
+  howpublished={\url{https://huggingface.co/0bserverx/Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF}}
+}
+```
+
+## Research use and responsibility
+
+This repository is intended for legitimate research and controlled evaluation, including interpretability, alignment and refusal-behavior analysis, red-team testing, and robustness work. It is not a ready-made production safety layer. If you deploy the model or expose it to other users, you are responsible for adding suitable access controls, moderation, monitoring, and abuse prevention.
+
+Use of these files is subject to the Apache License 2.0 and all applicable laws. You are responsible for how you operate the model and for outputs produced in your environment. To the extent permitted by law, the repository maintainers and upstream authors accept no liability for misuse or resulting harm. Generated outputs are not statements or endorsements by the maintainers, upstream creators, or their organizations.
diff --git a/RVN-BF16-mtp.gguf b/RVN-BF16-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..f585df9fc3bfb94c44b712911ef90b0e17056d2c
--- /dev/null
+++ b/RVN-BF16-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10813b480bd630d8c1148ccdccc9faf70afbf72cb6d46b8cf865c99112bef88f
+size 54259602656
diff --git a/RVN-BF16.gguf b/RVN-BF16.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..6a5d65d86329fa6eece6be1d3695be504c8f02e2
--- /dev/null
+++ b/RVN-BF16.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:461fa227cb3d0bb49a3e8e45ec2b650f9a2e593c51a92144d5ac819f0ee8f37d
+size 53808281888
diff --git a/RVN-F16-mtp.gguf b/RVN-F16-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..76e1fbc5c80767b6dcd89da8f7a65583ea59875c
--- /dev/null
+++ b/RVN-F16-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b08c52b2168f104aeeef9aaf88f6148f14986ccf03d733e4ddb8bb6daa17784
+size 54259602656
diff --git a/RVN-F16.gguf b/RVN-F16.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..94769336fe698ffcaa445f2a5718421c8a43f71a
--- /dev/null
+++ b/RVN-F16.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f20c72c23a6564211c5e44af35ea1277d697c009ce8a25ab28d100447e0f155a
+size 53808281888
diff --git a/RVN-IQ1_M-mtp.gguf b/RVN-IQ1_M-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..6bc8c82e0d0ad5f1f1b71e0cbdf5eb73224d8d55
--- /dev/null
+++ b/RVN-IQ1_M-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5edfe902abc5d2f83942f513f98df919129978b63ed36c8ba63ec4ad6ef8bc5b
+size 8082404768
diff --git a/RVN-IQ1_M-multilingual-mtp.gguf b/RVN-IQ1_M-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..a1fd84f8a3e571f333d0bd89f82c74c965a53fde
--- /dev/null
+++ b/RVN-IQ1_M-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04b045b94d02355478865224c7612417f489a36f0d6681450001ad5e5389cd4c
+size 8082404832
diff --git a/RVN-IQ1_M-multilingual.gguf b/RVN-IQ1_M-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..11c132498126d3ea86b659b06c73c8bfcccccc6b
--- /dev/null
+++ b/RVN-IQ1_M-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94d02f07eae45f0d71b56e3800ede1d28b9e18b4ab7685f552069a32836039dc
+size 7631084064
diff --git a/RVN-IQ1_M.gguf b/RVN-IQ1_M.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0588f6f43b1497ec87c0d509acf679f8ba44b4b2
--- /dev/null
+++ b/RVN-IQ1_M.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01a709e622165dddc275fb19c12444e20c6ecae18c068725db7d1aa2ee5fa45e
+size 7631084032
diff --git a/RVN-IQ1_S-mtp.gguf b/RVN-IQ1_S-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..958b72ac76cc97f6a0c7cae07b9b4c8f807a98b3
--- /dev/null
+++ b/RVN-IQ1_S-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77e292d92da4657f28ab1b8bfefdbadb1552d3e6daa17ff35e9dacce6e9ded59
+size 7601145248
diff --git a/RVN-IQ1_S-multilingual-mtp.gguf b/RVN-IQ1_S-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..a56f75e12598a615c1482b9c0cfc22214e43c37b
--- /dev/null
+++ b/RVN-IQ1_S-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:912cd2111d20ceab8273966667b79884d1194bd401e182336dac4e2bdf2b5ade
+size 7601145312
diff --git a/RVN-IQ1_S-multilingual.gguf b/RVN-IQ1_S-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..6fc2a2e7ad913a8c62e444dc7585c9dd417cdc97
--- /dev/null
+++ b/RVN-IQ1_S-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14331414fbf959c440d168ea39a005a8a0d8775b853e15f9e1d9df758f479737
+size 7149824544
diff --git a/RVN-IQ1_S.gguf b/RVN-IQ1_S.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..434ba5b3fe0bc905f5e9ba7a89fa21992f3af82e
--- /dev/null
+++ b/RVN-IQ1_S.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6718c15c056efa65755b9db664925425ad5111fabbf87397b77e82d9c223eee3
+size 7149824512
diff --git a/RVN-IQ2_M-mtp.gguf b/RVN-IQ2_M-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..e3f1b4bde3a5c558ffe40c614258f1fdc6815370
--- /dev/null
+++ b/RVN-IQ2_M-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc5e31bb27ba69ce4a0b0e92b71c4400a509578acc26b5944a64db6c9f2606c0
+size 10455913888
diff --git a/RVN-IQ2_M-multilingual-mtp.gguf b/RVN-IQ2_M-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..df7968e0bcd6a93b81712ec57f2ea101690ba30e
--- /dev/null
+++ b/RVN-IQ2_M-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:949967fd36fab7dd29f1a886a0e82d9566456e17b6bd1cb9da01983c55a73501
+size 10455913952
diff --git a/RVN-IQ2_M-multilingual.gguf b/RVN-IQ2_M-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..84e2590c88074df675f6e49af9400ede5e32fe87
--- /dev/null
+++ b/RVN-IQ2_M-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0fb26b1bd6fdf9c3eb84cbd97ce9405221bfce979cbf3c00ceaee6848682214
+size 10004593184
diff --git a/RVN-IQ2_M.gguf b/RVN-IQ2_M.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..264698d0c3189d44be761cdcb2d457c8ea6f178b
--- /dev/null
+++ b/RVN-IQ2_M.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a531c26993bb9ed91e299a025006915dc343285dc6dfabfd7d477370e42eb189
+size 10004593152
diff --git a/RVN-IQ2_S-mtp.gguf b/RVN-IQ2_S-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..fe1b74dd80aec60c00c5583190d4a1ddd215f6d4
--- /dev/null
+++ b/RVN-IQ2_S-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16256ace1d398695150ac971470062e64e6a992ebe6ee05492ef0e7af628dff7
+size 9814234560
diff --git a/RVN-IQ2_S-multilingual-mtp.gguf b/RVN-IQ2_S-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..3621567f4da02f8f341f873ca660569ae7401d5f
--- /dev/null
+++ b/RVN-IQ2_S-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6a7a567e4e789b9391996a7da59d2b11850fc79d167f178575847d5a186d767
+size 9814234592
diff --git a/RVN-IQ2_S-multilingual.gguf b/RVN-IQ2_S-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0d5ce0f1db2a34c06889d498e85c6adf8291e24e
--- /dev/null
+++ b/RVN-IQ2_S-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88c39f3a7297f0f4d3d85a8d02fff40202b0f0d490444f77b2d57fb99a099b8a
+size 9362913824
diff --git a/RVN-IQ2_S.gguf b/RVN-IQ2_S.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..bce264ecfa6163637a8c27a895d6147718b4f12e
--- /dev/null
+++ b/RVN-IQ2_S.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d0973eabea77748e02f114d74f149a74f4d878a09c56d47a3dd0bb9af36fada
+size 9362913792
diff --git a/RVN-IQ2_XS-mtp.gguf b/RVN-IQ2_XS-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..adf84e2655e13eede85df3c2e4188ea7dbf1338d
--- /dev/null
+++ b/RVN-IQ2_XS-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1107edd76991d5faee85781ec20df9511a2a41c19142158e7d14642300fafcf
+size 9541911968
diff --git a/RVN-IQ2_XS-multilingual-mtp.gguf b/RVN-IQ2_XS-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..09384b43e055396ef639eedb877b7b95f74088f8
--- /dev/null
+++ b/RVN-IQ2_XS-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ecdd5d7192cc146078baf601d085eed25acbd6784d9313ad3f955594217cf13
+size 9541912032
diff --git a/RVN-IQ2_XS-multilingual.gguf b/RVN-IQ2_XS-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..1a1c8f0cf7223a2a28d40540f0ef1fee43ba74e4
--- /dev/null
+++ b/RVN-IQ2_XS-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e7995899191c94875b3962443320452cd092608c06caae2b02e84a27bf2ae61
+size 9090591264
diff --git a/RVN-IQ2_XS.gguf b/RVN-IQ2_XS.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..14ce401620c813643da2dcc5bc30c27347bba0d8
--- /dev/null
+++ b/RVN-IQ2_XS.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:991858c2e6ed47347afbf5a35e47180ec7defd2d7b5e2ffbd705e4163241ea0e
+size 9090591232
diff --git a/RVN-IQ2_XXS-mtp.gguf b/RVN-IQ2_XXS-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..5099e80d3426d7fe94760d3bb834bcaa9d664b24
--- /dev/null
+++ b/RVN-IQ2_XXS-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:323b78d7ac87ccb546d885c7dd7f01da79dc97335bdd93e4ca4cf34b1de6d384
+size 8884503968
diff --git a/RVN-IQ2_XXS-multilingual-mtp.gguf b/RVN-IQ2_XXS-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..141fdba818a52adf6d360df0b1e5201886499ec4
--- /dev/null
+++ b/RVN-IQ2_XXS-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ce488ea9b2ae93216b318df449e0870cf155df27657137a6e0b0d300674700a
+size 8884504032
diff --git a/RVN-IQ2_XXS-multilingual.gguf b/RVN-IQ2_XXS-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..5563974ccf3d997b22fdc6424019fe2e51cfccc8
--- /dev/null
+++ b/RVN-IQ2_XXS-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08a157f354947a47df0aeca69f80d2477419341ab56ed1f5a2eb2f38bc994abd
+size 8433183264
diff --git a/RVN-IQ2_XXS.gguf b/RVN-IQ2_XXS.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..35221f61ce483c9924e315353e6ec3c0dcc20caf
--- /dev/null
+++ b/RVN-IQ2_XXS.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36578db18a17e726ff2de9bb459474d7a32d3822a35429ef026015335a57683f
+size 8433183232
diff --git a/RVN-IQ3_M-mtp.gguf b/RVN-IQ3_M-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..da3e62eee045126ebf5a692ed4b9f4cbbaa33c3a
--- /dev/null
+++ b/RVN-IQ3_M-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49b3e9cf17ea9e7de224bb1f4e60930798c99ba07bc2b6765c44f5e27c83304b
+size 13032195520
diff --git a/RVN-IQ3_M-multilingual-mtp.gguf b/RVN-IQ3_M-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..c1b8a7684d738b4461d021479a4643788d6fe4b6
--- /dev/null
+++ b/RVN-IQ3_M-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6deae4e6a0883fcf9a0e440902ef05a10b7a029d2717346c53b01c157c8a83db
+size 13032195552
diff --git a/RVN-IQ3_M-multilingual.gguf b/RVN-IQ3_M-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..b95a5fcbb8b509a22b16adbbc9f9b955f393bd5f
--- /dev/null
+++ b/RVN-IQ3_M-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0d060e9f0625e6eaeb40ce35c3344d6d6b8be2231f6f2a44eb19cc380fbf9fa
+size 12580874784
diff --git a/RVN-IQ3_M.gguf b/RVN-IQ3_M.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..8f83c0c42282545c068e371e4739808741408e05
--- /dev/null
+++ b/RVN-IQ3_M.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0f64d73d2ccfb5333a2e9dde9b079200d2a3e46f9ebaf19bc1a3cf14489d06b
+size 12580874752
diff --git a/RVN-IQ3_S-mtp.gguf b/RVN-IQ3_S-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..1acc1a9913623a7db89c7a1374271aad80835c50
--- /dev/null
+++ b/RVN-IQ3_S-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ce579dd7a81067c008089d292c36c451cf00796cb643ee9312d3dde10f7ca49
+size 12870649280
diff --git a/RVN-IQ3_S-multilingual-mtp.gguf b/RVN-IQ3_S-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..aaf25c9530bd55caf0cf7c6afd6ef2c4143c85ba
--- /dev/null
+++ b/RVN-IQ3_S-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48766d8382bf4e609402c51b2acc0b591f3d507e9eedc10e2ed7e3aa9030450b
+size 12870649312
diff --git a/RVN-IQ3_S-multilingual.gguf b/RVN-IQ3_S-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..7a1a0654343a121287c7dbdf2ded082cd53240b4
--- /dev/null
+++ b/RVN-IQ3_S-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8babfae3f1ee67f29e0c79154d71838a8f30f18bc7ada16b9e35bbf22dc5e30
+size 12419328544
diff --git a/RVN-IQ3_S.gguf b/RVN-IQ3_S.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..24419021215234d0dc3c13bb7f76617a85af40c2
--- /dev/null
+++ b/RVN-IQ3_S.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea1970358504bc9c0dd2fa8181820b3b6dc46951afe7d4473a06de34711f0606
+size 12419328512
diff --git a/RVN-IQ3_XS-mtp.gguf b/RVN-IQ3_XS-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..ca5d3b88069d74d575373bc2b2c401ac39548df2
--- /dev/null
+++ b/RVN-IQ3_XS-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abcb0d6e066209be5703353aa003ae2c393d57527e5c1d1468eafee32b9325c3
+size 12418450880
diff --git a/RVN-IQ3_XS-multilingual-mtp.gguf b/RVN-IQ3_XS-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..502b2ea9fab3abdce5100b2212c37689a8630cf8
--- /dev/null
+++ b/RVN-IQ3_XS-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f01165ccb5e818f44b9d719305cdd37066a2233e1f49d75a8f8557c2a4a46f4
+size 12418450912
diff --git a/RVN-IQ3_XS-multilingual.gguf b/RVN-IQ3_XS-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..71c60fed5d9019b9dcc7bdeb9f7cdf0518556a20
--- /dev/null
+++ b/RVN-IQ3_XS-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4fb2de73a9e61878affec30ba17a864b219f7aab21c26137ded4d7f159ecd00
+size 11967130144
diff --git a/RVN-IQ3_XS.gguf b/RVN-IQ3_XS.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..b0a96516e7ff1ab1a591406c8a9c1e2a69f0b114
--- /dev/null
+++ b/RVN-IQ3_XS.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fadbf639ade9862c426de3fbc0a5330557ba2a871041eef02f1d771701fbdb7
+size 11967130112
diff --git a/RVN-IQ3_XXS-mtp.gguf b/RVN-IQ3_XXS-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..3eb141dde37c0db03c0861a60dbba5d9804b1cf9
--- /dev/null
+++ b/RVN-IQ3_XXS-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99a6d6ae8ab3ecbefb0d1a0830d25f7b25f9168c83dfe8ab0266be023eb2e3f9
+size 11637691840
diff --git a/RVN-IQ3_XXS-multilingual-mtp.gguf b/RVN-IQ3_XXS-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..b9fe33c3aad55d6d49a4552ce9755567e7f743a0
--- /dev/null
+++ b/RVN-IQ3_XXS-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:338955bd8d67908afb4093b8a7b386dcfcd6eb6184a2171e5d6731c1d296abf6
+size 11637691872
diff --git a/RVN-IQ3_XXS-multilingual.gguf b/RVN-IQ3_XXS-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..1f70bc4c47a709b24216b6a5e3386957b306bd24
--- /dev/null
+++ b/RVN-IQ3_XXS-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4460c920ef12ece80f8a7e6896e9e3990cf90f477d7452c044f832a379287d9
+size 11186371104
diff --git a/RVN-IQ3_XXS.gguf b/RVN-IQ3_XXS.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..7bd79c2df9a79044769dafa4bd76098e977783fc
--- /dev/null
+++ b/RVN-IQ3_XXS.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e0e213d20eb289b081df65a403e118b9e5c5b932f75521703aaffc9d0b6f5e0
+size 11186371072
diff --git a/RVN-IQ4_NL-mtp.gguf b/RVN-IQ4_NL-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..8c2a8c43b41cee8211c7e0c6aeedbb3a52b5c141
--- /dev/null
+++ b/RVN-IQ4_NL-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dae10ccc69361055194867ad25ccb14b4043668d7d666b1028175e0c4bc72200
+size 16342377696
diff --git a/RVN-IQ4_NL-multilingual-mtp.gguf b/RVN-IQ4_NL-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..910ec1e38ea13b35a339386c5c7fb2e3f7f9c834
--- /dev/null
+++ b/RVN-IQ4_NL-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c443c8910022bc6a11ce0dce1a0157f46df10f7f8030119fb82603dbf836bb7
+size 16253248992
diff --git a/RVN-IQ4_NL-multilingual.gguf b/RVN-IQ4_NL-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..019d782caee02a7d08cef5275535ce5b76f6032a
--- /dev/null
+++ b/RVN-IQ4_NL-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7317d51ad903f9c547450070e3ce2aad74d7f27f6f02784e61446c2d01d0d8bf
+size 15801928224
diff --git a/RVN-IQ4_NL.gguf b/RVN-IQ4_NL.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..86fc555375de6ab2da00195e0842582a9dcdd771
--- /dev/null
+++ b/RVN-IQ4_NL.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38f8e785e6949f88a1027f9f1053df22e389f13a2deabf7877bf6bdef72adf3c
+size 15891056928
diff --git a/RVN-IQ4_XS-mtp.gguf b/RVN-IQ4_XS-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..d2dba080536cd45db22d46c2062a2eac38b3bd72
--- /dev/null
+++ b/RVN-IQ4_XS-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4052a5904dbf6615a8d6535913ecaa32862bddc9e5353f0a8557e2f95e4017a
+size 15645238496
diff --git a/RVN-IQ4_XS-multilingual-mtp.gguf b/RVN-IQ4_XS-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..fdbcaa0496890b3e90417ff2913e1c21ab84e8e5
--- /dev/null
+++ b/RVN-IQ4_XS-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28a03eb132c45a98d9493f0d51b5db57c38b376ec58e4892d3b441fd02128c86
+size 15533827552
diff --git a/RVN-IQ4_XS-multilingual.gguf b/RVN-IQ4_XS-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..f189b7816b81593c5180787881f8d0d7e8166a6a
--- /dev/null
+++ b/RVN-IQ4_XS-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b615856ba4a620db3be004ff5e5c46b3a6c5c3a17d247bef2cf6cbe1ca325aa4
+size 15082506784
diff --git a/RVN-IQ4_XS.gguf b/RVN-IQ4_XS.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..36fefc4c4910c2174eb675711c5932048898517b
--- /dev/null
+++ b/RVN-IQ4_XS.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3ba83d4f1750c270bc4fd57f0bb053ed52fb637686d5607a3c224f6a8f0de22
+size 15193917728
diff --git a/RVN-Q2_K-mtp.gguf b/RVN-Q2_K-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..350815da38246eea3910734803581de136c9893a
--- /dev/null
+++ b/RVN-Q2_K-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bed5048f3cb20ea7420f2d616635bbdf43f6c8e582f841f0d4f31b76fc7c19e
+size 11162985888
diff --git a/RVN-Q2_K-multilingual-mtp.gguf b/RVN-Q2_K-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0ae89a38aa7dfdf03d9c87ea4b939e4017a1029c
--- /dev/null
+++ b/RVN-Q2_K-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5a4fa4cc15a72013e6ab6fe1fea00b2b2eb96c9ed4317f5ec47ba85e86fdd4b
+size 11162985952
diff --git a/RVN-Q2_K-multilingual.gguf b/RVN-Q2_K-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..80998aae2a4a9a0c0a73620c6502c20423a3c29b
--- /dev/null
+++ b/RVN-Q2_K-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75b631012e4398d809e0c940c4056fbf94c1d0bebb8e179f9c0b9e6127c91cdf
+size 10711665184
diff --git a/RVN-Q2_K.gguf b/RVN-Q2_K.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..393910a9629aba39cc6970679360b13240dc7c77
--- /dev/null
+++ b/RVN-Q2_K.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3867c3e780fc73817f1c03bb33c869110747712d82d05e667f33e54f30f54259
+size 10711665152
diff --git a/RVN-Q2_K_S-mtp.gguf b/RVN-Q2_K_S-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..ec2e4a5288895b6dfd78c81289e15deabaac3f52
--- /dev/null
+++ b/RVN-Q2_K_S-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2883006f4e6d2b35b8c6ca23ee5a5dcefaf68be0bdaec319ea2b4df40bbfb02b
+size 10699646368
diff --git a/RVN-Q2_K_S-multilingual-mtp.gguf b/RVN-Q2_K_S-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..02099318d009b5d0750cfab96af3f02c9e317ef0
--- /dev/null
+++ b/RVN-Q2_K_S-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:698f70289fb99d9321fe8336f45ab8cd737a182af6295833249d2436b64c1cb0
+size 10699646432
diff --git a/RVN-Q2_K_S-multilingual.gguf b/RVN-Q2_K_S-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..49f4436c95f9d6fd50707e27bcbcd30f2b9fb6cb
--- /dev/null
+++ b/RVN-Q2_K_S-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcac0b9f8ebbd1ec1ec5e76c1d66b7d09cb89c50732cd99dad5ed1a252e3bd4e
+size 10248325664
diff --git a/RVN-Q2_K_S.gguf b/RVN-Q2_K_S.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..f968620497b686617acdabc05a097eb5b3b90fec
--- /dev/null
+++ b/RVN-Q2_K_S.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdd667b50c79b43f2e560ac27a975c49c3adf22674db551363b94ce5030c43f3
+size 10248325632
diff --git a/RVN-Q3_K_L-mtp.gguf b/RVN-Q3_K_L-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..aed468cb627b6c76e86092b36bd204b917edfa68
--- /dev/null
+++ b/RVN-Q3_K_L-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:458e4f4d4302e48933545e435d3d5e7444c85632702ab9d9be7f59d3a26a50fb
+size 14796096736
diff --git a/RVN-Q3_K_L-multilingual-mtp.gguf b/RVN-Q3_K_L-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..5c4d36357990b73b14f0f871da83c94420972947
--- /dev/null
+++ b/RVN-Q3_K_L-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0010af37eeff334c28cffde0f76c037bf4c093ba3ee7f3a74bb699a0f9ee8dff
+size 14796096992
diff --git a/RVN-Q3_K_L-multilingual.gguf b/RVN-Q3_K_L-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..2078a82c9466c36e65d0296c74a97a0a6b60a022
--- /dev/null
+++ b/RVN-Q3_K_L-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5129d2b03aef634dd42f57ddf08f06a30717c18069d1f7b47d91ad648abcacc
+size 14344776224
diff --git a/RVN-Q3_K_L.gguf b/RVN-Q3_K_L.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..f9c1fe8d07c44b4983fa6e29056193114dc293a5
--- /dev/null
+++ b/RVN-Q3_K_L.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b28a18047646dc079a528781b6b30b4a97e31fb246ddb7561d5f15ea8b6c9c96
+size 14344775968
diff --git a/RVN-Q3_K_M-mtp.gguf b/RVN-Q3_K_M-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..69fe9edee9ebe9e0aca4f0cbc30ad8a49d9c6dcf
--- /dev/null
+++ b/RVN-Q3_K_M-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:249b354e8e0a6984c903d5c099b9e9d054c1d5580b32e8dd818273dbacdb9a2c
+size 13752763616
diff --git a/RVN-Q3_K_M-multilingual-mtp.gguf b/RVN-Q3_K_M-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..088131160056b296c6255a11ca756bb56bf9a8d3
--- /dev/null
+++ b/RVN-Q3_K_M-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e268aa26c693011a2c1490fa3daf253a6af61f0c3c1e8ab40d6c0605ebd4742e
+size 13752763872
diff --git a/RVN-Q3_K_M-multilingual-vision.gguf b/RVN-Q3_K_M-multilingual-vision.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..e7def9eb5665b201543187453b950b3be7972c59
--- /dev/null
+++ b/RVN-Q3_K_M-multilingual-vision.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e54c0a8ddbe9c38b90cf8151c04bc27aceb78e33af954e2be5eb5a0359d49d2c
+size 16193856544
diff --git a/RVN-Q3_K_M-multilingual.gguf b/RVN-Q3_K_M-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0cf18eb7cb9a1e9438df2970e855f0a5dfefa072
--- /dev/null
+++ b/RVN-Q3_K_M-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be2c331e07a86a9243b49f6c17a25e46b5ff47e3b3a3852fa426606c87bfc563
+size 13301443104
diff --git a/RVN-Q3_K_M-vision.gguf b/RVN-Q3_K_M-vision.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..923ab9e69b766c5e93dee4aca3237fc73156dcb8
--- /dev/null
+++ b/RVN-Q3_K_M-vision.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02b2dd3542131722703e9863ff16b504f74b248fd01aeba873a88d5f8da857e3
+size 16193856288
diff --git a/RVN-Q3_K_M.gguf b/RVN-Q3_K_M.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..ea8559ce09ad38f25a39ba9ec72fdc77273b6ff4
--- /dev/null
+++ b/RVN-Q3_K_M.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08a875dc1cae0eec8ffc6ca529495f55d471d933588188ae49f925607937c3e1
+size 13301442848
diff --git a/RVN-Q3_K_S-mtp.gguf b/RVN-Q3_K_S-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..b2be7d8b5d3c5d30736f34716bc8a2d85f5f4412
--- /dev/null
+++ b/RVN-Q3_K_S-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ae918b94a083a0f7ba9ffa4c25994109d561bc2e9fa87dab6ce2f1a21b4bd9c
+size 12525274336
diff --git a/RVN-Q3_K_S-multilingual-mtp.gguf b/RVN-Q3_K_S-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..52577a77d7fdfc21ce6eb1751c8f3c770b02c35c
--- /dev/null
+++ b/RVN-Q3_K_S-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:221249e7b0649a336eb200adaf10357aacf7f3cde95d86165a8ff442070f1c2f
+size 12525274592
diff --git a/RVN-Q3_K_S-multilingual.gguf b/RVN-Q3_K_S-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..62ba472b59b6186e84ecd95cd374f0def6a0f8ec
--- /dev/null
+++ b/RVN-Q3_K_S-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:981e228e9fcf88366ea1e57b1f16ebb102483177bd36f3348fd188b833ab69a8
+size 12073953824
diff --git a/RVN-Q3_K_S.gguf b/RVN-Q3_K_S.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..c6bd5d0e3735c7c4cea09793ae0557f74b30464c
--- /dev/null
+++ b/RVN-Q3_K_S.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6419bd23d26ba9f3c2803779171c6e233dbed30631aa8d0350049c4011248a8
+size 12073953568
diff --git a/RVN-Q4_K_M-mtp.gguf b/RVN-Q4_K_M-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..3e85d2f2a078eca0f045bd04fa2859c396486e5d
--- /dev/null
+++ b/RVN-Q4_K_M-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5df52200763806fad5c01add7b1be13e9ef96dd1932a41226632693aac321b7b
+size 16998720736
diff --git a/RVN-Q4_K_M-multilingual-mtp.gguf b/RVN-Q4_K_M-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0ee7f86cf8274f47325637de207222538b269729
--- /dev/null
+++ b/RVN-Q4_K_M-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b20771570e85d1988aa3d1574698370a8b79f746eb02b96f25bf05c48f14b95
+size 16998720992
diff --git a/RVN-Q4_K_M-multilingual-vision.gguf b/RVN-Q4_K_M-multilingual-vision.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..757bbae6bf721cb9b1ce1a4ffa91e43af1cae7aa
--- /dev/null
+++ b/RVN-Q4_K_M-multilingual-vision.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b19bad318cea96ab21fc4d32a2bc4d05d3865e40e07a40ddb3c5122159232bb6
+size 18745293344
diff --git a/RVN-Q4_K_M-multilingual.gguf b/RVN-Q4_K_M-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..c0a07d32a1dfa0aec7a756267314404215731b37
--- /dev/null
+++ b/RVN-Q4_K_M-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ed40ccc8b8432f38b9a85d0ca67928167f5719e8c10bd299e56d34facaf6e61
+size 16547400224
diff --git a/RVN-Q4_K_M-vision.gguf b/RVN-Q4_K_M-vision.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..09b3c3be80054de7531f0fb241f3bfe2e7158f6b
--- /dev/null
+++ b/RVN-Q4_K_M-vision.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:331860c191da7601ef277a0f4f4fea9a7e44f5e422a0061dabe66605811c1304
+size 18745293088
diff --git a/RVN-Q4_K_M.gguf b/RVN-Q4_K_M.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..d9271a5646425605172dca5f2f0685426c855f33
--- /dev/null
+++ b/RVN-Q4_K_M.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c6789d3eac010a66dd14fc2e77b83b344d82d541937777cff4ea7b8e91d0a95
+size 16547399968
diff --git a/RVN-Q4_K_S-mtp.gguf b/RVN-Q4_K_S-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..9d573b0f01450337432dbcf9795befd04240154a
--- /dev/null
+++ b/RVN-Q4_K_S-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0730d3792f82502dbfbf9c812500cb873ee86bd77dc288485c0079b722b5e654
+size 16037635296
diff --git a/RVN-Q4_K_S-multilingual-mtp.gguf b/RVN-Q4_K_S-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..58055a6d548805d73d1e54c2a613810bcd019fea
--- /dev/null
+++ b/RVN-Q4_K_S-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:321a29c17b595e696f7c2e2c9e219336ac926179334219f70f51ddc4be609075
+size 16037635552
diff --git a/RVN-Q4_K_S-multilingual.gguf b/RVN-Q4_K_S-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..b2e6613201096d1b3797e60e83e412b0ab09ba9f
--- /dev/null
+++ b/RVN-Q4_K_S-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7f1b7220f7a05a5838d4f30d65799d38a7527f96ca9607019de6894c5e4861b
+size 15586314784
diff --git a/RVN-Q4_K_S.gguf b/RVN-Q4_K_S.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..4c7328e7cc2ca6b00260a47b2f5adc2df02e148b
--- /dev/null
+++ b/RVN-Q4_K_S.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3e1f1fd867d45c366e79773b106485dffa8efdb01bf3f75916f85a5c001f53b
+size 15586314528
diff --git a/RVN-Q5_K_M-mtp.gguf b/RVN-Q5_K_M-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..b9b804c167bac63f0fac3c84ebdcadc5585361ec
--- /dev/null
+++ b/RVN-Q5_K_M-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef6c307c53da1e0a577b27df0b636c2818880aabe5c132f423a404e36b391365
+size 19682419936
diff --git a/RVN-Q5_K_M-multilingual-mtp.gguf b/RVN-Q5_K_M-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..3943ce3a730c691d99d92a782a6c34ba20ef89bc
--- /dev/null
+++ b/RVN-Q5_K_M-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea26b22990fbb436d3192663fd50b5009385d2d01f86a049003f21c1badeb6a4
+size 19682420192
diff --git a/RVN-Q5_K_M-multilingual-vision.gguf b/RVN-Q5_K_M-multilingual-vision.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0fe696fe8c7fd9902040dd7c00789d83d072f354
--- /dev/null
+++ b/RVN-Q5_K_M-multilingual-vision.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:101b94796c6343f7ff9a361341d5bdc935d94126222ef37371c90fbd5a84d2ce
+size 21019351584
diff --git a/RVN-Q5_K_M-multilingual.gguf b/RVN-Q5_K_M-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0895a932b1c3c8e649b1d507076f315fe42c9fb1
--- /dev/null
+++ b/RVN-Q5_K_M-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1afe2e3fc7d18fcc3eabee0286b83cbc8d8680a4563722a5c489dc12519d8d02
+size 19231099424
diff --git a/RVN-Q5_K_M-vision.gguf b/RVN-Q5_K_M-vision.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..80a046027098eeb87479963e3103ccf0b4c99589
--- /dev/null
+++ b/RVN-Q5_K_M-vision.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8443359b3ba8fbcafe5ba698de3fbcf50cca4fd6ce33c1368c1c11e808a4016
+size 21019351328
diff --git a/RVN-Q5_K_M.gguf b/RVN-Q5_K_M.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..8120209ca61b8e24f7219c7c421f4d77ab5a1e7c
--- /dev/null
+++ b/RVN-Q5_K_M.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3996e017c40d7389a865453e2887d6d2952cbefe44c167bac47f0d9f454b9af0
+size 19231099168
diff --git a/RVN-Q5_K_S-mtp.gguf b/RVN-Q5_K_S-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..394852050505ec2afca50c97228515e5fd39335c
--- /dev/null
+++ b/RVN-Q5_K_S-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:119db236e256dbaa9dd247d18015caaa94d168d846cda5f888d3db3f0e9b2d67
+size 19130934496
diff --git a/RVN-Q5_K_S-multilingual-mtp.gguf b/RVN-Q5_K_S-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..181dd6c4edab59a194bce7d686f352e34c283d10
--- /dev/null
+++ b/RVN-Q5_K_S-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d18183d05dfd40a0837dd43b4ae956dcc73ea6daa618ecb7273229727840ce9
+size 19130934752
diff --git a/RVN-Q5_K_S-multilingual.gguf b/RVN-Q5_K_S-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..5619aac4dddcf86264f1927e46a3e8224c5defdf
--- /dev/null
+++ b/RVN-Q5_K_S-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71415ace64000b64613c999492bbdeaa74892fb4cd3b38aead6067393a3115fe
+size 18679613984
diff --git a/RVN-Q5_K_S.gguf b/RVN-Q5_K_S.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..92a8eda23116f0169ac9dd1d0c57f0f515158d0d
--- /dev/null
+++ b/RVN-Q5_K_S.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0440b346e0ccab79a5138884166b9e85b7d58f5ff68e6d3fe7ad8ec909884308
+size 18679613728
diff --git a/RVN-Q6_K-mtp.gguf b/RVN-Q6_K-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..e56a4a4c0d8d6ccb5218300ec1bebb8fa3c02bc7
--- /dev/null
+++ b/RVN-Q6_K-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e47a0e41992de4bed56a3395f6c7e1adb760a1875ed84f836f67d65b2f646ef
+size 22533850336
diff --git a/RVN-Q6_K-multilingual-mtp.gguf b/RVN-Q6_K-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..838c7f2a74ff696b72b2f0b04cc48aa3419abe52
--- /dev/null
+++ b/RVN-Q6_K-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1344d07425d73f0d1b8f36213910eae8fec21061b1a90835927e8485238f93a4
+size 22533850592
diff --git a/RVN-Q6_K-multilingual.gguf b/RVN-Q6_K-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..b3a3f6f7ea498179f56830d313ecda461e7670b0
--- /dev/null
+++ b/RVN-Q6_K-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:591be92b29db500eca2c75f8d0f292260426814418e50a8bf2f6e8a1ed29d138
+size 22082529824
diff --git a/RVN-Q6_K.gguf b/RVN-Q6_K.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..089863f8b12ff861fd440a79438f1c9aa1e2fa1a
--- /dev/null
+++ b/RVN-Q6_K.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:484fa7013ebc636fa217d559c30bebdbdca8a1be5d1296c86290acfa2fe8a5f2
+size 22082529568
diff --git a/RVN-Q8_0-mtp.gguf b/RVN-Q8_0-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..08dd29b227bd7cee96e7900bf32de290fede120f
--- /dev/null
+++ b/RVN-Q8_0-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea3101568bdebc7374ffa2bda80275862ad3d96a1279921e9b7fd649b82f3ab9
+size 29047084256
diff --git a/RVN-Q8_0-multilingual-mtp.gguf b/RVN-Q8_0-multilingual-mtp.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..0f5b2e5509d82657df1980a9affc5dc97b184e56
--- /dev/null
+++ b/RVN-Q8_0-multilingual-mtp.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3979ca0b400a091f60108906bd6a22907595e0dead3633bbda29b3400516f7bf
+size 29047084512
diff --git a/RVN-Q8_0-multilingual.gguf b/RVN-Q8_0-multilingual.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..7851a1d96504f0faadb406c19de38a6c6c45466d
--- /dev/null
+++ b/RVN-Q8_0-multilingual.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1e4974a6b70b58ceb3c255f64003ff69626963fe5ac2b0e122928eff96b737d
+size 28595763744
diff --git a/RVN-Q8_0.gguf b/RVN-Q8_0.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..e6d4e0034749a495c19db3d6b5268a262e897e7c
--- /dev/null
+++ b/RVN-Q8_0.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee66079819c13d6359ac0d70345ce39160386936a7f792cb668dfb705fab63c7
+size 28595763488
diff --git a/chat_template.jinja b/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..c0c686f9c38d70d179fb7b5f5aa7530bc913dda3
--- /dev/null
+++ b/chat_template.jinja
@@ -0,0 +1,170 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- set reasoning_instructions = '' %}
+{%- if enable_thinking is undefined or enable_thinking is true %}
+    {%- set resolved_reasoning_effort = reasoning_effort|default('xhigh') %}
+    {%- if resolved_reasoning_effort not in ('xhigh', 'medium', 'low') %}
+        {{- raise_exception('Unexpected reasoning effort ' ~ reasoning_effort ~ '. Supported types are xhigh (default), medium, and low.') }}
+    {%- endif %}
+    {%- if resolved_reasoning_effort == 'xhigh' %}
+        {%- set reasoning_instructions = 'Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.' %}
+    {%- elif resolved_reasoning_effort == 'low' %}
+        {%- set reasoning_instructions = 'Reasoning effort is set to low. Keep your thinking brief and focused, moving directly to the conclusion without unnecessary elaboration.' %}
+    {%- endif %}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {%- if reasoning_instructions %}
+        {{- reasoning_instructions + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '<|im_start|>system\n' + (reasoning_instructions + '\n\n' if reasoning_instructions else '')  + content + '<|im_end|>\n' }}
+        {%- elif reasoning_instructions %}
+            {{- '<|im_start|>system\n' + reasoning_instructions + '<|im_end|>\n' }}
+        {%- endif %}
+    {%- elif reasoning_instructions %}
+        {{- '<|im_start|>system\n' + reasoning_instructions + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if preserve_thinking is undefined or preserve_thinking is true or loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined and tool_call.arguments != '' %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/mmproj-Qwen3.8-27B-Q8_0.gguf b/mmproj-Qwen3.8-27B-Q8_0.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..1f53dcb74687a88208cdea802a0ebdb9efbdef5e
--- /dev/null
+++ b/mmproj-Qwen3.8-27B-Q8_0.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e968a6af97ce35d8971890b257b9b7edabf20ad91450501fa53162a19ee33eb
+size 629247008
diff --git a/mtp-RVN.gguf b/mtp-RVN.gguf
new file mode 100644
index 0000000000000000000000000000000000000000..d2ac284c3cab9c90a7a14a9df0cdb7a32d8bfc6b
--- /dev/null
+++ b/mtp-RVN.gguf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:655d00f1a1f77e8df0c0820c115347f952f64aff03e6a9daa3560ba4ed171e64
+size 1813125184
diff --git a/xet_speed_test.bin b/xet_speed_test.bin
new file mode 100644
index 0000000000000000000000000000000000000000..de5729ec81aa23c00401d39235f0d2a34a0ac1d1
--- /dev/null
+++ b/xet_speed_test.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:492becd1a45845adb1106b04fa01a1f4436fc1eaa0c3911244edf305d9c85b87
+size 104857600
diff --git a/xet_test_100mb.bin b/xet_test_100mb.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0e20887b3af294776d86231c12f2b78753a41f9f
--- /dev/null
+++ b/xet_test_100mb.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e867e2417a3783d784be70f9280c82f39d866f6f931421f690479e505951724c
+size 104857600