Update model card

README.md

@@ -14,7 +14,6 @@ tags:
 - qwen3
 - qwen3.5
 - qwen3.6
-- qwopus
 - claude-opus-distill
 - reasoning
 - vision
@@ -22,7 +21,6 @@ tags:
 - abliterated
 - refusal-ablated
 - uncensored
-- zerofuse
 - apple-silicon
 - mxfp4
 - ocp-mx
@@ -30,7 +28,6 @@ tags:
 - dflash
 - conversational
 base_model:
-- osmapi/osmQwopus-3.6-27B-V2-heretic-abliterated-uncensored
 - Jackrong/Qwopus3.6-27B-v2
 - Qwen/Qwen3.6-27B
 base_model_relation: quantized
@@ -38,13 +35,13 @@ pipeline_tag: image-text-to-text
 library_name: mlx
 ---
 
-# osmQwopus-3.6-27B-v2.1-uncensored-mxfp4
+# Qwen3.6-27B-v2.1-uncensored-mxfp4
 
 > ✅ **VISION PRESERVED.** Unlike most 4-bit quantization pipelines that degrade or strip the vision tower, this release keeps the full Qwen3.6-VL ViT and all projector weights in **fp16**. Use it as a real multimodal model at half the size of the 8-bit build.
 
-**MLX MXFP4 (OCP Microscaling FP4) quantization** of a ZeroFuse-abliterated Qwopus 3.6 27B v2 (the Jackrong Claude-Opus reasoning distill of Qwen 3.6 27B). Refusals reduced from **91/100 → 4/100** with KL drift of just **0.0176** (measured at BF16). Quantization adds only **+0.87% perplexity** with **92.4% greedy-token agreement** vs the BF16 reference. By the [osmAPI](https://osmAPI.com) research team and [TERV.Pro](https://terv.pro) student research team. Quantized with **[mlx-mtp](https://github.com/junainfinity/mlx-mtp)** — a vision-preserving MXFP4 quantizer.
+**MLX MXFP4 (OCP Microscaling FP4) quantization** of a ZeroFuse-abliterated Qwen 3.6 27B v2 (the Jackrong Claude-Opus reasoning distill of Qwen 3.6 27B). Refusals reduced from **91/100 → 4/100** with KL drift of just **0.0176** (measured at BF16). Quantization adds only **+0.87% perplexity** with **92.4% greedy-token agreement** vs the BF16 reference. By the Lemura Labs research team. Quantized with **[mlx-mtp](https://github.com/junainfinity/mlx-mtp)** — a vision-preserving MXFP4 quantizer.
 
-> Fastest absolute throughput of any osmQwopus MLX build — **27.7 tok/s vanilla** on M4 Max. Apple Silicon is memory-bandwidth-bound: halving the model fits more weights in fast SRAM, which matters more than the precision drop.
+> Fastest absolute throughput of any Qwen3.6-27B MLX build — **27.7 tok/s vanilla** on M4 Max. Apple Silicon is memory-bandwidth-bound: halving the model fits more weights in fast SRAM, which matters more than the precision drop.
 
 ---
 
@@ -57,33 +54,33 @@ library_name: mlx
 | **Scheme** | OCP Microscaling FP4 (MXFP4) — `mode=mxfp4`, `group_size=32` |
 | **Vision tower** | ✅ fp16 — all 333 ViT + projector weights at full precision |
 | **SSM params** | ✅ fp16 — `a_log`, `dt_bias`, `conv1d` kept at full precision |
-| **Refusal rate (ZeroFuse, n=100)** | **4/100** (vs vanilla Qwopus **91/100**) |
+| **Refusal rate (ZeroFuse, n=100)** | **4/100** (vs vanilla Qwen 3.6 **91/100**) |
 | **KL divergence vs BF16 (quant)** | **0.034 nats** mean · **JSD 0.009** (essentially identical) |
 | **Top-1 greedy agreement vs BF16** | **92.4%** — same argmax token at 9 in 10 positions |
 | **PPL increase vs BF16** | **+0.87%** — negligible quality loss |
-| **Throughput (M4 Max, vanilla)** | **27.7 tok/s** — fastest of all osmQwopus MLX builds |
+| **Throughput (M4 Max, vanilla)** | **27.7 tok/s** — fastest of all Qwen3.6-27B MLX builds |
 | **Throughput + DFlash (b=8)** | **29.3 tok/s** (1.06×) |
 | **Recommended RAM** | 20 GB+ Apple Silicon — fits M2 Pro / M3 Pro and above |
 | **Best for** | Fastest local inference; 16–32 GB Macs; vision + text |
-| **Released by** | [osmAPI](https://osmAPI.com) · [TERV.Pro](https://terv.pro) |
+| **Released by** | Lemura Labs |
 
 ---
 
-## 🎚️ All osmQwopus variants
+## 🎚️ All Qwen3.6-27B variants
 
-The full osmQwopus family from [osmAPI](https://osmAPI.com) — same ZeroFuse-abliterated weights (refusal 4/100, KL 0.0176), different quant schemes for different runtimes.
+The full Qwen3.6-27B family from Lemura Labs — same ZeroFuse-abliterated weights (refusal 4/100, KL 0.0176), different quant schemes for different runtimes.
 
 | Quant | Format | BPW | Disk | Vision | Runtime | Link |
 |---|---|---|---|---|---|---|
-| 8-bit | MLX | 8.50 | ~27 GB | ✅ native | mlx-vlm | [`…-8-bit-mlx`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-8-bit-mlx) |
-| 6-bit | MLX | 6.66 | ~21 GB | ✅ native | mlx-vlm | [`…-6-bit-mlx`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-6-bit-mlx) |
+| 8-bit | MLX | 8.50 | ~27 GB | ✅ native | mlx-vlm | [`…-8-bit-mlx`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-8-bit-mlx) |
+| 6-bit | MLX | 6.66 | ~21 GB | ✅ native | mlx-vlm | [`…-6-bit-mlx`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-6-bit-mlx) |
 | **MXFP4** **(this repo)** | MLX | **4.25** | **~14 GB** | ✅ **fp16 tower** | mlx-mtp / mlx-vlm | — *(you are here)* |
-| OptiQ 3.7bpw | MLX | ~3.7 | ~14 GB | ✅ ViT spliced | mlx-vlm | [`…-OptiQ-3.7bpw-mlx`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-OptiQ-3.7bpw-mlx) |
-| Q8_0 | GGUF | 8.50 | ~28 GB | ✅ via mmproj | llama.cpp | [`…-8-bit-GGUF`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-8-bit-GGUF) |
-| Q6_K | GGUF | ~6.56 | ~22 GB | ✅ via mmproj | llama.cpp | [`…-6-bit-GGUF`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-6-bit-GGUF) |
-| Q4_K_M | GGUF | ~4.92 | ~16 GB | ✅ via mmproj | llama.cpp | [`…-Q4_K_M-GGUF`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-Q4_K_M-GGUF) |
-| TQ3_4S | GGUF | 4.00 (~3.5 eff) | ~14 GB | ✅ via mmproj | [llama.cpp-tq3](https://github.com/turbo-tan/llama.cpp-tq3) | [`…-TQ3_4s-GGUF`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF) |
-| TQ3_1S | GGUF | 4.00 (~3.5 eff) | ~14 GB | ✅ via mmproj | [llama.cpp-tq3](https://github.com/turbo-tan/llama.cpp-tq3) | [`…-TQ3_1s-GGUF`](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-TQ3_1s-GGUF) |
+| OptiQ 3.7bpw | MLX | ~3.7 | ~14 GB | ✅ ViT spliced | mlx-vlm | [`…-OptiQ-3.7bpw-mlx`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-OptiQ-3.7bpw-mlx) |
+| Q8_0 | GGUF | 8.50 | ~28 GB | ✅ via mmproj | llama.cpp | [`…-8-bit-GGUF`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-8-bit-GGUF) |
+| Q6_K | GGUF | ~6.56 | ~22 GB | ✅ via mmproj | llama.cpp | [`…-6-bit-GGUF`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-6-bit-GGUF) |
+| Q4_K_M | GGUF | ~4.92 | ~16 GB | ✅ via mmproj | llama.cpp | [`…-Q4_K_M-GGUF`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-Q4_K_M-GGUF) |
+| TQ3_4S | GGUF | 4.00 (~3.5 eff) | ~14 GB | ✅ via mmproj | [llama.cpp-tq3](https://github.com/turbo-tan/llama.cpp-tq3) | [`…-TQ3_4s-GGUF`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF) |
+| TQ3_1S | GGUF | 4.00 (~3.5 eff) | ~14 GB | ✅ via mmproj | [llama.cpp-tq3](https://github.com/turbo-tan/llama.cpp-tq3) | [`…-TQ3_1s-GGUF`](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-TQ3_1s-GGUF) |
 
 > 👉 All variants share the same abliterated base weights — pick by your runtime (Apple Silicon → MLX; CUDA/CPU/cross-platform → GGUF) and your RAM budget. MXFP4 is the fastest MLX build on M4 Max; use 8-bit if you want the MTP head for native speculative decoding without a drafter.
 
@@ -92,29 +89,29 @@ The full osmQwopus family from [osmAPI](https://osmAPI.com) — same ZeroFuse-ab
 ## 🧬 Lineage
 
 ```
-Qwen/Qwen3.6-27B                              (Qwen Team — base multimodal pretrain)
-        │
-        ▼
-Jackrong/Qwopus3.6-27B-v2                     (Jackrong — Claude-Opus reasoning distill)
-        │
-        ▼
-ZeroFuse abliteration (TPE-50)          (osmAPI · TERV.Pro)
-   ├── 25 random startup trials
-   ├── 2 community priors (coder3101, wangzhang)
-   └── 23 TPE smart-sampling trials → best at trial 45
-        │
-        ▼
-osmapi/osmQwopus-3.6-27B-V2-zerofuse           (BF16 full-precision checkpoint)
-        │
-        ▼
-this repo — MLX MXFP4 quant                  (junainfinity/mlx-mtp · osmAPI · TERV.Pro)
+Qwen/Qwen3.6-27B (Qwen Team — base multimodal pretrain)
+ │
+ ▼
+Jackrong/Qwopus3.6-27B-v2 (Jackrong — Claude-Opus reasoning distill)
+ │
+ ▼
+ZeroFuse abliteration (TPE-50) (Lemura Labs · TERV.Pro)
+ ├── 25 random startup trials
+ ├── 2 community priors (coder3101, wangzhang)
+ └── 23 TPE smart-sampling trials → best at trial 45
+ │
+ ▼
+lemuralabs/Qwen3.6-27B-V2-zerofuse (BF16 full-precision checkpoint)
+ │
+ ▼
+this repo — MLX MXFP4 quant (junainfinity/mlx-mtp · Lemura Labs · TERV.Pro)
 ```
 
 Direct upstream links:
 
 * 🏛️ Foundation: [Qwen/Qwen3.6-27B](https://huggingface.co/Qwen/Qwen3.6-27B)
 * 🎓 Claude-Opus distill: [Jackrong/Qwopus3.6-27B-v2](https://huggingface.co/Jackrong/Qwopus3.6-27B-v2)
-* 🔓 Abliteration tool: [ZeroFuse](https://github.com/junainfinity/ZeroFuse) by [osmAPI](https://osmAPI.com)
+* 🔓 Abliteration tool: [ZeroFuse](https://github.com/junainfinity/ZeroFuse) by Lemura Labs
 * 🧮 Quantization tool: [mlx-mtp](https://github.com/junainfinity/mlx-mtp) — vision-preserving MXFP4 quantizer
 
 ---
@@ -137,11 +134,11 @@ ZeroFuse measures refusals on `mlabonne/harmful_behaviors` (100 hard red-team pr
 
 ## 🧪 Method
 
-**Abliteration** (inherited from the [BF16 source](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored)) — TPE-50 Optuna search on ZeroFuse, M4 Max 128 GB. Full method in that card.
+**Abliteration** (inherited from the [BF16 source](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored)) — TPE-50 Optuna search on ZeroFuse, M4 Max 128 GB. Full method in that card.
 
 **MXFP4 quantization** (this repo):
 
-1. **Source** — loaded `osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored` (51 GB BF16, 3 shards) into MLX on M4 Max.
+1. **Source** — loaded `lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored` (51 GB BF16, 3 shards) into MLX on M4 Max.
 2. **Layer audit** — `mlx_mtp.mxfp4_quantize` catalogued 1199 tensors: identified all LM linear projections (attention Q/K/V/O + MLP gate/up/down + lm_head) as MXFP4 candidates; flagged 196 vision-tower + projector tensors and 48 SSM-sensitive params (`a_log`, `dt_bias`, `conv1d`) for fp16 preservation.
 3. **Quantization** — LM linears → MXFP4 (OCP MX E2M1, `group_size=32`) using `mlx.core.quantize`. Vision + SSM tensors written verbatim at fp16. 15 MTP head weights absent from source — quantizer set `mtp_num_hidden_layers: 0` in output config to prevent strict-load failure.
 4. **Output** — 3 safetensor shards, 14 GB total. Embedded `mlx_mtp` config block records format, vision_fp16=True, mtp_preserved=False.
@@ -179,12 +176,12 @@ git clone https://github.com/junainfinity/mlx-mtp && pip install -e ./mlx-mtp
 ```python
 from mlx_mtp.engine import load_model, vanilla_generate
 
-model, processor, config = load_model("osmapi/osmQwopus-3.6-27B-v2.1-uncensored-mxfp4")
+model, processor, config = load_model("lemuralabs/Qwen3.6-27B-v2.1-uncensored-mxfp4")
 
 result = vanilla_generate(
-    model, processor, config,
-    text="Write a Python function that implements merge sort:",
-    max_tokens=512,
+ model, processor, config,
+ text="Write a Python function that implements merge sort:",
+ max_tokens=512,
 )
 print(result["text"])
 print(f"{result['tps']:.1f} tok/s")
@@ -196,14 +193,14 @@ print(f"{result['tps']:.1f} tok/s")
 from mlx_mtp.engine import load_model
 from mlx_mtp.dflash import load_dflash_drafter, dflash_generate
 
-model, processor, config = load_model("osmapi/osmQwopus-3.6-27B-v2.1-uncensored-mxfp4")
-drafter, _              = load_dflash_drafter("z-lab/Qwen3.6-27B-DFlash")
+model, processor, config = load_model("lemuralabs/Qwen3.6-27B-v2.1-uncensored-mxfp4")
+drafter, _ = load_dflash_drafter("z-lab/Qwen3.6-27B-DFlash")
 
 result = dflash_generate(
-    model, processor, config, drafter,
-    text="Explain the key differences between TCP and UDP:",
-    max_tokens=512,
-    draft_block_size=8,   # block 8 > 16 > 32 on mxfp4 (model is fast; keep overhead low)
+ model, processor, config, drafter,
+ text="Explain the key differences between TCP and UDP:",
+ max_tokens=512,
+ draft_block_size=8, # block 8 > 16 > 32 on mxfp4 (model is fast; keep overhead low)
 )
 print(result["text"])
 print(f"{result['tps']:.1f} tok/s")
@@ -220,8 +217,8 @@ from mlx_vlm import load, generate
 from mlx_vlm.prompt_utils import apply_chat_template
 from mlx_vlm.utils import load_config
 
-model, processor = load("osmapi/osmQwopus-3.6-27B-v2.1-uncensored-mxfp4")
-config = load_config("osmapi/osmQwopus-3.6-27B-v2.1-uncensored-mxfp4")
+model, processor = load("lemuralabs/Qwen3.6-27B-v2.1-uncensored-mxfp4")
+config = load_config("lemuralabs/Qwen3.6-27B-v2.1-uncensored-mxfp4")
 
 messages = [{"role": "user", "content": "Explain the difference between SSM and softmax attention in 3 sentences."}]
 prompt = apply_chat_template(processor, config, messages)
@@ -232,10 +229,10 @@ print(generate(model, processor, prompt, max_tokens=400, verbose=True))
 
 ```python
 out = generate(
-    model, processor,
-    prompt=apply_chat_template(processor, config, [{"role": "user", "content": "Describe this picture."}], num_images=1),
-    image=["./photo.jpg"],
-    max_tokens=400, verbose=True,
+ model, processor,
+ prompt=apply_chat_template(processor, config, [{"role": "user", "content": "Describe this picture."}], num_images=1),
+ image=["./photo.jpg"],
+ max_tokens=400, verbose=True,
 )
 print(out)
 ```
@@ -244,33 +241,33 @@ print(out)
 
 ```bash
 pip install mlx-omni-server
-mlx-omni-server  # serves on http://127.0.0.1:10240
+mlx-omni-server # serves on http://127.0.0.1:10240
 ```
 
 ```bash
 curl -s http://127.0.0.1:10240/v1/chat/completions \
-  -H "Content-Type: application/json" \
-  -d '{
-    "model": "osmapi/osmQwopus-3.6-27B-v2.1-uncensored-mxfp4",
-    "messages": [
-      {"role": "system", "content": "Be brief and direct."},
-      {"role": "user", "content": "Write a 3-line haiku about a heisenbug."}
-    ],
-    "max_tokens": 200
-  }' | jq -r '.choices[0].message.content'
+ -H "Content-Type: application/json" \
+ -d '{
+ "model": "lemuralabs/Qwen3.6-27B-v2.1-uncensored-mxfp4",
+ "messages": [
+ {"role": "system", "content": "Be brief and direct."},
+ {"role": "user", "content": "Write a 3-line haiku about a heisenbug."}
+ ],
+ "max_tokens": 200
+ }' | jq -r '.choices[0].message.content'
 ```
 
 ---
 
 ## 🧪 Quantization details
 
-* **Source weights**: `osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored` — BF16, 3 shards, ~51 GB.
+* **Source weights**: `lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored` — BF16, 3 shards, ~51 GB.
 * **Quantization scheme**: OCP Microscaling FP4 (MXFP4) — 4-bit E2M1 mantissa (2 exponent bits, 1 mantissa bit) + shared E8M0 scale per group of 32 elements. Hardware-accelerated on Apple M-series via MLX's native MX support.
 * **Group size**: 32.
 * **Effective bits/weight**: 4 + 8/32 = **4.25 bpw** (8-bit E8M0 scale shared across 32 weights).
 * **Vision tower**: preserved verbatim in fp16 — 333 vision weights (ViT encoder, vision projector, image-token embeddings, deepstack indexes). Pixel/patch features degrade visibly at 4-bit; keeping them fp16 is free (vision = minority of total params).
 * **SSM-sensitive params**: preserved in fp16 — 48 params (`a_log`, `dt_bias`, `conv1d`). Qwen3.5's hybrid-SSM recurrent state is numerically sensitive; fp16 costs nothing at scale.
-* **MTP head**: disabled (`mtp_num_hidden_layers: 0`). The BF16 source had 0 MTP weights; quantizer detects this and clears the config to prevent strict-load failure. For native MTP speculative decoding, use the [8-bit build](https://huggingface.co/osmapi/osmQwopus-3.6-27B-V2-abliterated-uncensored-8-bit-mlx).
+* **MTP head**: disabled (`mtp_num_hidden_layers: 0`). The BF16 source had 0 MTP weights; quantizer detects this and clears the config to prevent strict-load failure. For native MTP speculative decoding, use the [8-bit build](https://huggingface.co/lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-8-bit-mlx).
 * **Quantizer**: [`mlx_mtp.mxfp4_quantize`](https://github.com/junainfinity/mlx-mtp/blob/main/mlx_mtp/mxfp4_quantize.py).
 
 ### Architecture notes
@@ -283,7 +280,7 @@ Qwen 3.6 27B uses a **hybrid attention stack** — 3 linear-attention (GatedDelt
 
 * **Uncensored.** Refusal directions were surgically removed; this model will answer prompts the parent would refuse. Use responsibly and within applicable law. The release is provided for safety research, red-teaming, and creative/educational use cases.
 * **Identity preserved.** The model still self-identifies as Qwen (developed by Alibaba's Tongyi Lab) — abliteration does not rewrite factual self-knowledge.
-* **Heavy chain-of-thought.** Qwopus inherits Claude-Opus's verbose reasoning style. For terse answers, use a system prompt like `"Be brief and direct. Skip your reasoning."`.
+* **Heavy chain-of-thought.** Qwen 3.6 inherits Claude-Opus's verbose reasoning style. For terse answers, use a system prompt like `"Be brief and direct. Skip your reasoning."`.
 
 ---
 
@@ -295,12 +292,11 @@ Qwen 3.6 27B uses a **hybrid attention stack** — 3 linear-attention (GatedDelt
 
 **Quantization & release**
 
-[osmAPI research team](https://osmAPI.com) — *"OpenRouter of India"*
-[TERV.Pro student research team](https://terv.pro)
+Lemura Labs
 
 **Claude-Opus reasoning distill**
 
-[Jackrong](https://huggingface.co/Jackrong) — author of [Qwopus3.6-27B-v2](https://huggingface.co/Jackrong/Qwopus3.6-27B-v2)
+[Jackrong](https://huggingface.co/Jackrong) — author of [Qwen 3.6 27B-v2](https://huggingface.co/Jackrong/Qwopus3.6-27B-v2)
 
 **Foundation model**
 
@@ -308,7 +304,7 @@ Qwen 3.6 27B uses a **hybrid attention stack** — 3 linear-attention (GatedDelt
 
 **Abliteration toolkit**
 
-[ZeroFuse](https://github.com/junainfinity/ZeroFuse) by [osmAPI](https://osmAPI.com) — Optuna-driven refusal-direction ablation with KL guardrails.
+[ZeroFuse](https://github.com/junainfinity/ZeroFuse) by Lemura Labs — Optuna-driven refusal-direction ablation with KL guardrails.
 
 **Community priors that seeded the TPE search**
 
@@ -322,11 +318,11 @@ Qwen 3.6 27B uses a **hybrid attention stack** — 3 linear-attention (GatedDelt
 
 ## 📜 License
 
-Apache-2.0, inherited from the foundation (Qwen3.6-27B) and the distill (Qwopus3.6-27B-v2) upstream.
+Apache-2.0, inherited from the foundation (Qwen3.6-27B) and the distill (Qwen 3.6 27B-v2) upstream.
 
 ---
 
-Need a hosted endpoint, custom quant, or larger-scale inference? [osmAPI](https://osmAPI.com) — multi-provider LLM routing for the Indian developer ecosystem.
+Need a hosted endpoint, custom quant, or larger-scale inference? Lemura Labs — multi-provider LLM routing for the Indian developer ecosystem.
 
 ---
 
@@ -336,10 +332,10 @@ This MLX build supports **block-diffusion speculative decoding** via **[DFlash](
 
 ```
 Apple M4 Max · 256 tok greedy · 2026-06-11
-  MXFP4  vanilla          27.7 tok/s   1.00×
-  MXFP4  + DFlash b=8     29.3 tok/s   1.06×  ← optimal
-  MXFP4  + DFlash b=16    26.2 tok/s   0.95×
-  MXFP4  + DFlash b=32    24.2 tok/s   0.87×
+ MXFP4 vanilla 27.7 tok/s 1.00×
+ MXFP4 + DFlash b=8 29.3 tok/s 1.06× ← optimal
+ MXFP4 + DFlash b=16 26.2 tok/s 0.95×
+ MXFP4 + DFlash b=32 24.2 tok/s 0.87×
 ```
 
 **Why the modest speedup?** DFlash's drafter overhead becomes a larger fraction of total time as baseline throughput rises. The 8-bit build sees ~3× gains because it starts slower; mxfp4 is already bandwidth-limited in a faster regime. If raw throughput is the priority, mxfp4 vanilla at 27.7 tok/s already beats oQ8 + DFlash (best = 23 tok/s).
@@ -350,12 +346,12 @@ Apple M4 Max · 256 tok greedy · 2026-06-11
 from mlx_mtp.engine import load_model
 from mlx_mtp.dflash import load_dflash_drafter, dflash_generate
 
-model, processor, config = load_model("osmapi/osmQwopus-3.6-27B-v2.1-uncensored-mxfp4")
+model, processor, config = load_model("lemuralabs/Qwen3.6-27B-v2.1-uncensored-mxfp4")
 drafter, _ = load_dflash_drafter("z-lab/Qwen3.6-27B-DFlash")
 
 out = dflash_generate(model, processor, config, drafter,
-                      text="Write a merge function for two sorted lists in Python.",
-                      max_tokens=256, draft_block_size=8)
+ text="Write a merge function for two sorted lists in Python.",
+ max_tokens=256, draft_block_size=8)
 print(out["text"])
 ```
 
@@ -363,9 +359,9 @@ print(out["text"])
 
 ```bash
 python3 -m mlx_vlm generate \
-  --model osmapi/osmQwopus-3.6-27B-v2.1-uncensored-mxfp4 \
-  --draft-model z-lab/Qwen3.6-27B-DFlash --draft-kind dflash \
-  --prompt "Write a merge function for two sorted lists in Python." --max-tokens 256
+ --model lemuralabs/Qwen3.6-27B-v2.1-uncensored-mxfp4 \
+ --draft-model z-lab/Qwen3.6-27B-DFlash --draft-kind dflash \
+ --prompt "Write a merge function for two sorted lists in Python." --max-tokens 256
 ```
 
 - Requires access to the gated drafter [`z-lab/Qwen3.6-27B-DFlash`](https://huggingface.co/z-lab/Qwen3.6-27B-DFlash) (one-click "Agree and access").