---
license: apache-2.0
base_model:
- Jackrong/Qwopus3.6-27B-v2
- Jackrong/Qwopus3.6-27B-Coder-MTP-GGUF
- Jackrong/Qwopus3.6-35B-A3B-Coder
- Jackrong/Qwopus3.6-35B-A3B-Coder-MTP-GGUF
base_model_relation: quantized
datasets:
- Jackrong/Claude-opus-4.6-TraceInversion-9000x
- Jackrong/Claude-opus-4.7-TraceInversion-5000x
- lambda/hermes-agent-reasoning-traces
language:
- en
- zh
- es
- ru
- ja
library_name: llama.cpp
pipeline_tag: image-text-to-text
tags:
- qwopus3.6
- qwen3.6
- qwen35
- 27b
- 35b
- model-size-27b
- model-size-35b
- dense
- moe
- coder
- agentic-coding
- tool-calling
- gguf
- mtp
- rocm
- rocmfp4
- rocmfpx
- moequality
- chadrock
- llama.cpp
- amd
- ryzen-ai-max-395
- strix-halo
- vision
- multimodal
---

![CHADROCK3.6 Coder ROCmFPX / ROCmFP4 MTP](assets/chadrockcoder.png)

# CHADROCK3.6 Coder ROCmFPX / ROCmFP4 MTP

This repo contains the CHADROCK3.6 Coder releases for AMD Ryzen AI Max+ 395 / Strix Halo systems:

- `CHADROCK3.6-27B-Coder-MTP-ROCmFP4-STRIX_LEAN.gguf`: the original dense 27B ROCmFP4/MTP Coder lane.
- `CHADROCK3.6-35B-A3B-Coder-MTP-ROCmFPX-MoEQuality-7.08BPW.gguf`: the additive 35B A3B MoEQuality ROCmFPX/MTP Coder lane.

The existing 27B ROCmFP4 lane is preserved. The 35B MoEQuality file is added as a second download lane for users who want the larger A3B Coder behavior with a higher-quality ROCmFPX tensor mix.

CHADROCK3.6 Coder uses Qwopus3.6 Coder lineage, then converts the source models into Charlie's AMD-focused ROCmFP4 / ROCmFPX runtime formats. The public release names and artifact names are Chadrock names, while Qwopus stays explicit in lineage, base model metadata, and credits.

The 27B file is a compact 14 GB GGUF for local agentic coding, repository work, tool-use style prompts, and long-context experiments. The 35B A3B MoEQuality file is a larger 30 GB GGUF aimed at better tool-use/coder behavior while keeping draft-MTP serving available on unified-memory AMD hardware.

These GGUFs will **not run correctly with stock llama.cpp**. They need a pinned [`ciru-ai/ROCmFPX`](https://github.com/ciru-ai/ROCmFPX/tree/7aa484a2f0a504dc612a3d74a068024f3e6d6353) runner because the files use ROCmFP4 / ROCmFPX tensor types that upstream llama.cpp does not currently understand.

The model file is already provided here. You do **not** need to rebuild or quantize the model. Build the custom llama server once, download the files, and run the profile below.

## Why This Build Exists

CHADROCK3.6 27B Coder is the Strix-focused Chadrock release of a dense agentic coding model lineage. It is intended for coding, tool use, debugging, structured developer workflows, and runtime experimentation on AMD hardware. Chadrock adds the AMD runtime piece:

- ROCmFP4 Strix Lean tensor recipe
- native draft-MTP serving
- AMD ROCm/HIP backend path
- 262K context target
- q4_0 KV cache profile for long local sessions
- optional vision projector companion file

This release is best treated as a model/runtime pairing for Strix Halo rather than a generic GGUF quant.

## Model Lineage

```text
Qwen/Qwen3.6-27B
  -> Jackrong/Qwopus3.6-27B-v2
       datasets:
         - Jackrong/Claude-opus-4.6-TraceInversion-9000x
         - Jackrong/Claude-opus-4.7-TraceInversion-5000x
  -> Jackrong/Qwopus3.6-27B-Coder
       adds:
         - lambda/hermes-agent-reasoning-traces
         - agentic coding and tool-use SFT
  -> Jackrong/Qwopus3.6-27B-Coder-MTP-GGUF
  -> jcbtc/chadrock3.6-27b-coder-rocmfp4-mtp

Qwen/Qwen3.6-35B-A3B
  -> Jackrong/Qwopus3.6-35B-A3B-Coder
  -> Jackrong/Qwopus3.6-35B-A3B-Coder-MTP-GGUF
  -> CHADROCK3.6-35B-A3B-Coder-MTP-ROCmFPX-MoEQuality-7.08BPW.gguf
```

In plain terms: Qwen provides the foundation models, Jackrong's Qwopus lines add Trace Inversion and coder/tool-use training, the upstream MTP GGUFs provide the MTP sources, and this release converts those lines into Strix-focused Chadrock runtime formats.

## Technical Metadata

### 27B ROCmFP4 Lane

| Field | Value |
| --- | --- |
| model size | `27B` dense |
| architecture | `qwen35` |
| local runtime format | ROCmFP4 Chadrock GGUF |
| direct upstream/source GGUF | `Jackrong/Qwopus3.6-27B-Coder-MTP-GGUF` |
| upstream behavior lineage | `Jackrong/Qwopus3.6-27B-v2` plus coder SFT |
| local profile | `qwopus3.6-27b-coder-mtp-chadrock-rocmfp4-strix-lean` |
| context target | `262144` tokens |
| draft mode | `draft-mtp`, `n_max=4`, `p_split=0.10` |
| intended hardware | AMD Ryzen AI Max+ 395 / Strix Halo |

### 35B A3B ROCmFPX MoEQuality Lane

| Field | Value |
| --- | --- |
| model size | `35B A3B` MoE |
| architecture | `qwen35moe` |
| local runtime format | ROCmFPX MoEQuality GGUF |
| direct upstream/source GGUF | `Jackrong/Qwopus3.6-35B-A3B-Coder-MTP-GGUF` |
| source revision | `f629cb8638d27e92c09361c8d9c91389c0fbc712` |
| source artifact | `Qwopus3.6-35B-A3B-Coder-MTP-Q8_0.gguf` |
| local profile | `qwopus36-35b-coder-mtp-rocmfpx-moequality-708bpw-hermes64k-froggeric-template` |
| output BPW | `7.08 BPW` |
| context target used for Tool Eval | `65536` tokens |
| draft mode | `draft-mtp`, `n_max=3`, `p_min=0.25`, `p_split=0.10` |
| target KV / draft KV | `q8_0 / q8_0` target, `f16 / f16` draft |
| chat template | Froggeric Qwen fixed chat template, SHA256 `27d22ab352efbb63cdcc379cc58924f16b2949931e6f185b959f8930efc9520b` |
| intended hardware | AMD Ryzen AI Max+ 395 / Strix Halo |

## Local Benchmark Notes

All numbers below were measured locally on AMD Ryzen AI Max+ 395 / Strix Halo.

### Tool Eval Full 69 - 35B A3B ROCmFPX MoEQuality

The 35B A3B MoEQuality lane was run through the local Tool Eval full 69 deterministic tool-use suite with the raw llama.cpp endpoint, `temperature=0`, `seed=42`, `parallel=1`, `--no-think`, and `--structured-response-format json_object`.

| Metric | Result |
| --- | ---: |
| final score | `72` |
| points | `100 / 138` |
| scenarios | `40 pass / 20 partial / 9 fail` |
| median turn time | `3708.1 ms` |
| generated throughput during eval | `21.14 tok/s` |

The run artifacts were audited for the earlier structured-output harness failure mode. No HTTP 400, sampler initialization, or backend schema-support failure signatures were found in the JSON, progress log, or generated report. The structured-output cases below are scored model behavior, not backend failures.

| Category | Label | Score |
| --- | --- | ---: |
| A | Tool Selection | `6 / 6 = 100%` |
| B | Parameter Precision | `6 / 6 = 100%` |
| C | Multi-Step Chains | `8 / 8 = 100%` |
| D | Restraint & Refusal | `5 / 6 = 83%` |
| E | Error Recovery | `5 / 6 = 83%` |
| F | Localization | `6 / 6 = 100%` |
| G | Structured Reasoning | `2 / 6 = 33%` |
| H | Instruction Following | `8 / 10 = 80%` |
| I | Context & State | `14 / 20 = 70%` |
| J | Code Patterns | `4 / 6 = 67%` |
| K | Safety & Boundaries | `18 / 26 = 69%` |
| L | Toolset Scale | `5 / 8 = 62%` |
| M | Autonomous Planning | `4 / 6 = 67%` |
| N | Creative Composition | `3 / 6 = 50%` |
| O | Structured Output | `6 / 12 = 50%` |

### BigCodeBench Hard Instruct

| Run | Result |
| --- | ---: |
| `bigcodebench-hard-instruct`, calibrated | `48/148 = 32.43% pass@1` |

The scored run used the local profile `qwopus3.6-27b-coder-mtp-chadrock-rocmfp4-strix-lean` in the June 13 full coding benchmark folder. That local profile name records source lineage and build path; the public release name is CHADROCK3.6 27B Coder.

### Apples-to-Apples Q5_K_M Comparison

The cleanest decode-speed comparison is the same CHADROCK CLI guard run against the upstream Qwopus3.6 27B Coder MTP `Q5_K_M` GGUF and this Chadrock ROCmFP4 build, using the same prompts, runtime build, machine, and MTP guard harness.

| Guard row | Upstream Q5_K_M decode | Chadrock ROCmFP4 decode | Decode uplift |
| --- | ---: | ---: | ---: |
| short arithmetic prompt | `17.8 tok/s` | `29.5 tok/s` | `1.66x` |
| sustained regression-guard prompt | `13.2 tok/s` | `22.6 tok/s` | `1.71x` |

Prompt processing is better represented by the no-cache long-context sweep below, where Chadrock ROCmFP4 measured `315.97 tok/s` at 4K prompt tokens and `142.00 tok/s` at 130K prompt tokens.

### Long-Context Sweep

The no-cache forced context sweep generated 512 tokens at each context length:

| Prompt tokens | Prompt speed | Decode speed | Draft accepted |
| ---: | ---: | ---: | ---: |
| `4,131` | `315.97 tok/s` | `21.25 tok/s` | `314/779` |
| `8,227` | `308.66 tok/s` | `21.82 tok/s` | `329/728` |
| `16,419` | `286.62 tok/s` | `21.64 tok/s` | `344/666` |
| `32,803` | `251.76 tok/s` | `17.35 tok/s` | `335/701` |
| `65,571` | `201.49 tok/s` | `12.51 tok/s` | `329/726` |
| `130,467` | `142.00 tok/s` | `7.08 tok/s` | `305/823` |

These are local server measurements, not universal llama.cpp claims. Throughput depends heavily on driver version, clocks, prompt shape, KV cache settings, and MTP acceptance.

## Best Settings / Advanced Setup

For the pinned runner build, copy-paste build commands, request-level speculative controls, and the latest Chadrock ROCmFP4 reproduction notes, use the advanced Ciru setup page:

**https://llm.ciru.ai/chadrock-rocmfpx/**

The current pinned runner build is:

```text
ciru-ai/ROCmFPX commit: 7aa484a2f0a504dc612a3d74a068024f3e6d6353
historical score tag: chadrock-rocmfp4-mtp-scores-20260621
```

For this published `CHADROCK3.6-27B-Coder` GGUF, the validated card profile is:

```text
backend: ROCm0 target + ROCm0 draft
context: 262144
batch / ubatch: 512 / 512
target KV: q4_0 / q4_0
draft KV: q4_0 / q4_0
MTP: draft-mtp, n_max=4, n_min=0, p_min=0.0, p_split=0.10
serving: one slot, metrics on, no mmap
sampler: temperature=1.0, top_p=0.95, top_k=20, reasoning off
```

For the added `CHADROCK3.6-35B-A3B-Coder-MTP-ROCmFPX-MoEQuality-7.08BPW` GGUF, the Tool Eval profile was:

```text
backend: Vulkan0 target + Vulkan0 draft
context: 65536
batch / ubatch: 2048 / 512
target KV: q8_0 / q8_0
draft KV: f16 / f16
MTP: draft-mtp, n_max=3, n_min=0, p_min=0.25, p_split=0.10
serving: one slot, metrics on, no context shift, text-only with --no-mmproj
sampler: temperature=0, top_p=0.95, top_k=20, seed=123, reasoning off
chat template: Froggeric Qwen fixed chat template
```

Use the advanced page if you are testing the newer request-level ROCmFPX runner
or comparing against the separate Qwable 5 27B Coder ROCmFP4 speed lane. The
settings above are the best published-card settings for the actual GGUF in this
repo.

## Run With llama-server

### 27B ROCmFP4 Lane

Build Charlie's custom llama.cpp once, download this GGUF and the projector file, then run:

```bash
HSA_OVERRIDE_GFX_VERSION=11.5.1 \
GGML_HIP_ENABLE_UNIFIED_MEMORY=1 \
/path/to/rocmfp4-llama/build-strix-rocmfp4/bin/llama-server \
  -m CHADROCK3.6-27B-Coder-MTP-ROCmFP4-STRIX_LEAN.gguf \
  --mmproj mmproj-F32.mmproj \
  --alias chadrock3.6-27b-coder \
  --host 127.0.0.1 \
  --port 8080 \
  --jinja \
  -c 262144 \
  -ngl 999 \
  -fa on \
  -dev ROCm0 \
  -b 512 \
  -ub 512 \
  -t 16 \
  -tb 32 \
  -ctk q4_0 \
  -ctv q4_0 \
  --ctx-checkpoints 0 \

  --checkpoint-every-n-tokens -1 \

  --spec-type draft-mtp \
  --spec-draft-device ROCm0 \
  --spec-draft-ngl all \
  --spec-draft-type-k q4_0 \
  --spec-draft-type-v q4_0 \
  --spec-draft-n-max 4 \
  --spec-draft-n-min 0 \
  --spec-draft-p-min 0.0 \
  --spec-draft-p-split 0.10 \
  --parallel 1 \
  --metrics \
  --no-mmap
```

Use `--parallel 1` for this MTP profile. Multi-slot serving changes draft-MTP behavior and is not the intended configuration.

For text-only use, you may omit `--mmproj`.

For vision use, keep `mmproj-F32.mmproj` beside the main GGUF, but run with MTP off. In practice, that means using the vision projector and removing the `--spec-*` draft-MTP flags from the command.

The projector is a GGUF-format projector file with a `.mmproj` repo extension so Hugging Face's GGUF metadata badge tracks the 27B language model rather than the smaller projector.

### 35B A3B ROCmFPX MoEQuality Lane

For the 35B MoEQuality file, use a ROCmFPX runner with Vulkan support and the same Froggeric Qwen fixed chat template used for the local Tool Eval run:

```bash
/path/to/rocmfpx-vulkan/bin/llama-server \
  -m CHADROCK3.6-35B-A3B-Coder-MTP-ROCmFPX-MoEQuality-7.08BPW.gguf \
  --alias chadrock3.6-35b-a3b-coder-moequality \
  --host 127.0.0.1 \
  --port 8080 \
  --jinja \
  -c 65536 \
  --reasoning off \
  --reasoning-format none \
  -sm none \
  -ngl 999 \
  -fa on \
  -b 2048 \
  -ub 512 \
  --no-context-shift \
  -dev Vulkan0 \
  --chat-template-file /path/to/froggeric-qwen-fixed-chat-template.jinja \
  -t 16 \
  -tb 32 \
  -ctk q8_0 \
  -ctv q8_0 \
  --spec-type draft-mtp \
  --spec-draft-device Vulkan0 \
  --spec-draft-ngl all \
  --spec-draft-type-k f16 \
  --spec-draft-type-v f16 \
  --spec-draft-n-max 3 \
  --spec-draft-n-min 0 \
  --spec-draft-p-min 0.25 \
  --spec-draft-p-split 0.10 \
  --no-spec-draft-backend-sampling \
  --parallel 1 \
  --temp 0 \
  --top-p 0.95 \
  --top-k 20 \
  --seed 123 \
  --metrics
```

For text-only serving, omit the projector. If you use the included `mmproj-CHADROCK3.6-35B-A3B-Coder-MTP-F32.mmproj` projector, validate your vision path separately and run with MTP off unless your local runner supports that combination.

## Build The Required llama.cpp

```bash
git clone https://github.com/ciru-ai/ROCmFPX.git
cd ROCmFPX
git checkout 7aa484a2f0a504dc612a3d74a068024f3e6d6353
env JOBS=16 scripts/build-strix-rocmfp4-mtp.sh llama-server llama-bench
```

The server binary will be here:

```text
build-strix-rocmfp4/bin/llama-server
```

## About ROCmFP4 / Chadrock

Charlie's ROCmFP4 method adds AMD-focused GGUF tensor formats and backend paths to llama.cpp.

ROCmFP4 is not stock Q4, MXFP4, or NVFP4. It uses custom 4-bit tensor layouts, Codebook10 values, finite unsigned E4M3 scale semantics, tensor-aware Strix presets, ROCm/HIP kernels, Vulkan support, and MTP regression guards.

Why it matters: Strix Halo has a large unified-memory pool, but good local serving still depends on memory bandwidth, tensor layout, KV traffic, and draft-token acceptance. Chadrock is built for that exact hardware shape.

## Files

| File | Size | SHA256 |
| --- | ---: | --- |
| `CHADROCK3.6-27B-Coder-MTP-ROCmFP4-STRIX_LEAN.gguf` | `14 GB` | `9536a6d9d56708a6b9e94cde00bde59a1788834ce58fa3b37eabfa8626e325d0` |
| `mmproj-F32.mmproj` | `889 MB` | `32f7ea0600c07272547da401d460f8abbd980f3a57b69d6df87be0e2505e0b9c` |
| `CHADROCK3.6-35B-A3B-Coder-MTP-ROCmFPX-MoEQuality-7.08BPW.gguf` | `30 GB` | `db23284e3c7ddf088392d3b89fcec8dc1b4e1830846f7670f0fc48e749a2cf2a` |
| `mmproj-CHADROCK3.6-35B-A3B-Coder-MTP-F32.mmproj` | `1.7 GB` | `5c82c8095717b39f29c88ebfec3607a10307785b1e14a87744603d6c582cd497` |

## Credits

- **[Qwen](https://huggingface.co/Qwen)**: `Qwen/Qwen3.6-27B` base model family.
- **[Jackrong](https://huggingface.co/Jackrong)**: Qwopus3.6 v2, Qwopus3.6 27B Coder, Trace Inversion datasets, coder/tool-use SFT, and the MTP GGUF source.
- **[lambda](https://huggingface.co/lambda)**: `lambda/hermes-agent-reasoning-traces`, included by the upstream coder release.
- **charlie12345 / [@Italianclownz](https://x.com/Italianclownz)**: ROCmFP4 llama.cpp fork, Strix Halo build path, and AMD-focused MTP runtime work.

## Notes

This is an experimental AMD ROCmFP4/MTP build. It is intended for local evaluation, coding workflows, and runtime experimentation on compatible AMD hardware.