Tess-4-27B-ROCmFP4_FAST-GGUF

Tess-4-27B ROCmFP4_FAST — GGUF

ROCmFP4_FAST quant of migtissera/Tess-4-27B (Apache 2.0), produced via charlie12345/ROCmFPX's ROCm fork. Tess-4-27B is a Qwen3.6-27B fine-tune. Weights: 14 GB, ~3.8 bpw ROCmFP4_FAST preset. Companion TQ3_4S quant for Blackwell CUDA is in a separate repo.

File

NOT a stock llama.cpp quant

ROCmFP4_FAST is a custom weight format unique to charlie12345/ROCmFPX. Stock llama.cpp and the nixpkgs llama-cpp package will exit with unknown quantization at load time. Use the llama-server/llama-cli from the ROCmFPX fork.

Scope of these benchmarks — read this first

These numbers are a light baseline, not a thorough ROCmFPX evaluation. The mesh's bench framework is built for production agent workload regression-detection on the local stack, not for the kind of multi-axis sweep that upstream quant maintainers typically publish. Specifically:

• Harness scope is bounded. Numbers come from production inference throughput monitoring and the mesh's 6-test regression suite (mesh_eval). That's a regression suite, not a quality benchmark.
• Sample sizes are small. Single GPU, single rep. None are powered for statistical significance.
• No perplexity / wikitext / MMLU / GSM8K. The mesh's stack isn't a quality benchmark.
• Single GPU class (RDNA4 16 GB). All measurements on an AMD RX 9060 XT 16 GB (gfx1200, ROCm 7.x). No Blackwell, no CDNA, no multi-GPU. Cross-hardware generalization is NOT implied.
• No human eval. "Loads, runs, doesn't break the agent stack" is not a quality verdict on this specific quant.

What this IS good for: a quick signal that the quant (a) loads on RDNA4, (b) runs at production-usable throughput, (c) doesn't break the mesh's agent tool-calling. What this is NOT good for: claiming "this is the best quant of this model," reproducing academic benchmark results, or substituting for upstream's validation work.

For a rigorous view, see migtissera/Tess-4-27B (parent model) and charlie12345/ROCmFPX (quantizer).

Quick start

# Build charlie12345/ROCmFPX (ROCm fork)
git clone https://github.com/charlie12345/ROCmFPX
cd ROCmFPX
mkdir build && cd build
cmake .. -DGGML_HIP=ON -DAMDGPU_TARGETS=gfx1200
make -j$(nproc)

# Serve
llama-server \
  -m Tess-4-27B-ROCmFP4_FAST.gguf \
  --host 0.0.0.0 --port 8081 \
  -ngl 99 -c 65536 -t 12 \
  -ctk q4_0 -ctv q4_0 \
  -fa on --cache-ram 0 --no-cache-prompt \
  -np 1 --batch-size 512 --ubatch-size 128 \
  --jinja --metrics -rea off

Reproduce the quant

# Requires the ROCmFPX fork and the F16 source GGUF
llama-quantize --allow-requantize tess-4-27b-f16.gguf \
  Tess-4-27B-ROCmFP4_FAST.gguf Q4_0_ROCMFP4_FAST

Files in this repo

What's NOT in this repo (caveats)

• Stock llama.cpp will not load this file. ROCmFP4_FAST is unique to charlie12345/ROCmFPX.
• No CUDA / non-AMD GPU bench. All measurements are RDNA4 (gfx1200). The companion TQ3_4S quant for Blackwell is in a separate repo.
• No AEON bench data yet. This quant was tested in production for regression but has not run the full AEON suite. The companion TQ3_4S quant has AEON results (0.560 mean).
• No quality benchmark (perplexity, MMLU, GSM8K). The custom 4-bit quant works on the mesh's regression tests; whether it's "the best ROCmFPX quant" needs upstream validation.
• No MTP / speculative-decode bench. Tess-4 has native MTP support; this quant was tested without MTP.
• No vision/multimodal test. This variant is text-only.

Provenance

• Source model: migtissera/Tess-4-27B — Qwen3.6-27B fine-tune
• Source model license: Apache 2.0
• Quantizer: charlie12345/ROCmFPX
• Quantizer license: MIT
• Build hardware: AMD RX 9060 XT 16 GB (RDNA4, gfx1200), ROCm 7.x, NixOS 25.11
• Bench harness: Mesh production stack (mesh_eval regression)

License

• Tess-4-27B is Apache 2.0 (per its HF model card).
• charlie12345/ROCmFPX is MIT.
• The GGUF in this repo is a derivative of the Apache 2.0-licensed parent, produced with the MIT-licensed quantizer. The Apache 2.0 license is preserved.