Qwen3.8-27B NVFP4 for NInfer

This model card is the version-controlled source for neroued/Qwen3.8-27B-nvfp4-NInfer.

The repository contains a mixed NVFP4/FP8 representation of Qwen3.8-27B. It combines the official BF16 checkpoint with the fixed packed Text weights from unsloth/Qwen3.8-27B-NVFP4 in the native NInfer .ninfer artifact format. The artifact is intended only for NInfer; it is not a Transformers checkpoint, Safetensors distribution, or GGUF file.

The artifact uses the Qwen3.5 Dense architecture. Text layers 0–55 use NVFP4 MLP weights, while the token embedding, attention input/output projections, GDN Q/K/V/Z and output projections, full output head, and Text layers 56–63 MLP weights use row-scaled FP8. Control weights use BF16, with separate MTP, Vision and DFlash2 weights.

Artifact

The file contains Text, Vision, MTP, DFlash2, the optimized proposal head and frontend resources. Vision and speculative weights are loaded only when selected at startup. Source-derived NVFP4 and FP8 words are preserved without decode and requantization; only the official BF16 token embedding is encoded locally as row-scaled FP8.

Verify a downloaded file with:

printf '%s  %s\n' \
  '74d2c57145e6ff11d1d2faa79594477f9bc903a611af1fb20218189fbbb77d82' \
  'qwen3_8_27b_nvfp4.ninfer' | sha256sum --check

This release includes the complete DFlash2 companion weights from z-lab/Qwen3.8-27B-DFlash2 at revision 50307d4c4cde6860d4eee73e2547cd786fe8e8a4. Select --spec dflash2 --draft-tokens 7 --proposal-head optimized; draft counts 1..15 are supported. Use the current runtime for DFlash2. Existing performance and evaluation tables retain their stated MTP configurations and revisions.

Requirements

• Current NInfer checkout, built from source;
• 64-bit Linux;
• NVIDIA GeForce RTX 5090 (sm_120a);
• Build dependencies from the build guide.

Already have the official v2 file? Upgrade it locally without downloading the weights again.

NInfer does not provide an install target or packaged binary. See the repository README for source-build dependencies.

Download and run a CLI example

hf download neroued/Qwen3.8-27B-nvfp4-NInfer \
  qwen3_8_27b_nvfp4.ninfer \
  --local-dir models

./build/rtx5090-release/apps/ninfer models/qwen3_8_27b_nvfp4.ninfer \
  --prompt "Explain prefill and decode in three sentences." \
  --max-context 32768 \
  --max-tokens 8192 \
  --kv-dtype fp8 \
  --spec mtp --draft-tokens 3 \
  --proposal-head optimized

For images, videos, and structured chat history, see the CLI guide.

Start a local server

From the NInfer repository root, copy the RTX 5090 deployment example and set model.path to the downloaded artifact:

cp examples/configs/rtx5090-qwen3_8-27b-nvfp4.toml ninfer.local.toml
# Edit model.path for your artifact.
./build/rtx5090-release/apps/ninfer-serve --config ninfer.local.toml

The example enables Vision, FP8 KV, CUDA Graph decode, MTP3 and the optimized proposal head, with two active requests, a 1,024-token prefill chunk and 256 MiB of memory headroom. Each request has a 240,000-token logical ceiling. KV histories and continuation state share the Device Context byte capacity selected at startup. The log reports the pool size and estimated token capacity. Requests allocate as they advance; resource pressure can reclaim retained prefixes or pause requests for later restoration/replay. See the startup reference for capacity, Host cache and deployment controls.

See the HTTP serving guide for the API surface and the resource scheduling reference for cache and admission semantics.

Supported use

The artifact supports:

• text generation in thinking and non-thinking modes;
• image, multi-image, video, and mixed multimodal messages;
• MTP speculative decoding with draft windows from one to five;
• DFlash2 with draft windows from one to fifteen using the included DFlash2 companion weights (--spec dflash2 --draft-tokens 7, optionally --proposal-head optimized);
• BF16, INT8, FP8, NVFP4, and K8V4 KV cache;
• CUDA Graph decode and compatible-prefix reuse;
• startup-bounded small-scale concurrent serving with true batched decode;
• the NInfer CLI;
• OpenAI Responses Core, OpenAI Chat Completions, and Anthropic Messages serving.

Performance

Measured on September 28–29, 2026 with NInfer revision 7f6aafed, one RTX 5090, driver 617.14, and CUDA 13.4 compile/runtime/driver API. These serving runs use FP8 E4M3 row-256 KV, CUDA Graphs, a 1,024-token prefill chunk, disabled prefix reuse, and temperature 0.6 / top-p 0.95 / top-k 20 / min-p 0 / presence penalty 1.0 / frequency penalty 0. MTP0 has a 262,144-token context ceiling; MTP3 uses 131,072 tokens per request, three draft tokens, the optimized proposal head, and automatic shared KV capacity.

Concurrent MTP=3 corpus makespan

Each C is one complete 75-request corpus, with three reasoning and twelve cross-scenario fixtures, five seeds per fixture, and a fixed shuffled send order. Makespan includes prefill, decode, admission waits, transitions, and drain; actual output lengths vary.

All 300 requests completed without request, CUDA, or allocation errors. At C=1/2/4/8, automatic KV capacity is 131,072 / 262,144 / 253,632 / 225,024 tokens. C=8 has average batch 3.52 and up to five waiting requests in the sampled intervals. The resident MTP3 weights occupy 19.729 GiB, and the workspace arena is 243.3 MiB.

Long-context serving (MTP disabled)

Values are arithmetic mean ± sample standard deviation over five fixed seeds per fixture.

MTP=3 single-request long-reasoning decode

The C=1 corpus supplies these phase statistics, with five samples per reasoning fixture.

MTP=3 single-request cross-scenario decode

Each category pools three fixtures × five seeds. Values are mean ± sample standard deviation.

All five AIME 15 samples reach the 65,536-token output budget. The C=1 corpus contains 40 stop-token and 35 output-limit results; all are retained in the statistics. These measurements do not score answer accuracy or task completion.

The full results and reproduction commands also cover DFlash2 K=7, MTP3 decode saturation, and completion outcomes.

Evaluation

The artifact was evaluated through NInfer's OpenAI-compatible serving route with thinking enabled, MTP=3, and INT8 group-64 KV. EvalScope 1.9.0 used 0-shot prompts, rule-based scoring, and one sample per problem with temperature 1.0, top-p 0.95, top-k 20, presence penalty 0.0, and seed 42. The text suite ran at a 252,928-token context limit; the multimodal suite ran with --vision at a 81,920-token limit.

All 1,723 configured samples completed and were scored. IFBench additionally reports 80.50% instruction-level strict, 80.33% prompt-level loose, and 83.50% instruction-level loose. These are single-sample results, not pass@k.

The official Qwen3.8-27B BF16 figures come from the upstream model card; its sampling settings and IFBench metric level are not stated there, so the last column is not a same-protocol comparison. The NVFP4 deltas stay within ±2.5 points on the four overlapping benchmarks, and the upstream card reports no AIME results.

Limits

• This artifact targets one RTX 5090, with one to eight active requests per Engine.
• KV and continuation state share the startup Context pool. Usable context depends on the selected KV type, enabled components, concurrency and available memory.
• NInfer returns generated tool calls to the client; it does not execute them.

Provenance

The artifact identity, summarized object inventory, and conversion provenance are published in artifact-manifest.json. The exact storage contract is maintained in the v3 container reference.

License

This NInfer artifact is distributed under the Apache License 2.0. The Qwen3.8-27B base repository and the quantized source repository are also licensed under Apache-2.0. Users remain responsible for complying with the license and applicable laws.