Qwen3.8-27B-NVFP4

Qwen3.8-27B-NVFP4

Model Overview

• Model Architecture: Qwen3_5ForConditionalGeneration
  • Input: Text / Image
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP4 and FP8
  • Activation quantization: FP4 and FP8
• Release Date: 2026-09-21
• Version: 2.0
• Model Developers: RedHatAI

This model is an updated quantized version of Qwen/Qwen3.8-27B, using a mixed-precision FP4/FP8 scheme with an unquantized language-model head and updated quantization scales. See Evaluation for accuracy results.

Model Optimizations

This model was produced by applying mixed-precision quantization to Qwen/Qwen3.8-27B. MLP projections are quantized to FP4, attention projections and the final MLP layers are quantized to FP8, and the KV cache is quantized to FP8, while the language-model head is kept in full precision to preserve output quality. The quantization scales were updated by calibrating on a 512-sample subset of the perfectblend dataset with a recipe that combines AWQ and GPTQ.

Only the weights and activations of the linear operators within the transformer blocks are quantized using LLM Compressor. The checkpoint is ~24.7 GB on disk (versus ~54 GB in BF16), reducing disk size and GPU memory requirements by roughly 70%.

Deployment

vLLM Serving

vllm serve RedHatAI/Qwen3.8-27B-NVFP4 \
  --reasoning-parser qwen3 \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_xml \
  --speculative-config '{"method":"mtp","num_speculative_tokens":3}' \

For optimal peformance, consider using the DSpark draft model RedHatAI/Qwen3.8-27B-speculator.dspark for speculative decoding, shown below.

vllm serve RedHatAI/Qwen3.8-27B-NVFP4 \
  --reasoning-parser qwen3 \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_xml \
  --speculative-config '{"model":"RedHatAI/Qwen3.8-27B-speculator.dspark","num_speculative_tokens":7,"method":"dspark"}'

See <b>Performance Evaluation</b> below for further details.

Creation

This model was created by applying LLM Compressor with calibration samples from perfectblend, as presented in the code snippet below.

<details>

from compressed_tensors.quantization.quant_scheme import (
    FP8_DYNAMIC,
    NVFP4,
    QuantizationScheme,
)
from transformers import AutoProcessor, Qwen3_5ForConditionalGeneration

from llmcompressor import oneshot
from llmcompressor.modifiers.gptq import GPTQModifier
from llmcompressor.modifiers.transform.awq import AWQModifier
from llmcompressor.utils import load_context

MODEL_ID = "Qwen/Qwen3.8-27B"

# Load model.
with load_context(Qwen3_5ForConditionalGeneration):
    model = Qwen3_5ForConditionalGeneration.from_pretrained(MODEL_ID)
processor = AutoProcessor.from_pretrained(MODEL_ID)


recipe = [
    AWQModifier(duo_scaling="both"),
    GPTQModifier(
        config_groups={
            "attention": QuantizationScheme(
                targets=[
                    r"re:.*self_attn\.(q|k|v|o)_proj$",
                    r"re:.*linear_attn\.(in_proj_qkv|in_proj_z|out_proj)$",
                    r"re:.*layers\.(56|57|58|59|60|61|62|63)\.mlp\..*(gate|up|down)_proj$",
                ],
                **FP8_DYNAMIC,
            ),
            "mlp": QuantizationScheme(
                targets=[r"re:.*mlp\..*(gate|up|down)_proj$"],
                **NVFP4,
            ),
        },
        ignore=[
            "re:visual.*",
            "re:model.visual.*",
            "re:.*lm_head",
        ],
        kv_cache_scheme={
            "num_bits": 8,
            "type": "float",
            "symmetric": True,
            "strategy": "tensor",
            "dynamic": False,
            "observer": "static_minmax",
        },
    ),
]

# Apply quantization.
oneshot(
    model=model,
    processor=processor,
    recipe=recipe,
    dataset="perfectblend",
    splits="train[:512]",
    max_seq_length=4096,
    num_calibration_samples=512,
    moe_calibrate_all_experts=True,
)

# Save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-NVFP4-GPTQ-AWQ"
model.save_pretrained(SAVE_DIR)
processor.save_pretrained(SAVE_DIR)

</details>

Evaluation

This model was evaluated on GSM8K Platinum, MATH-500, AIME 2025, GPQA Diamond, and IFEval using lm-evaluation-harness and lighteval, and on SWE Bench using Inspect AI, served with vLLM (OpenAI-compatible API). Evaluations were run on 1x B200 GPU.

Accuracy

Recovery vs. BF16 baseline

<table> <thead> <tr> <th>Category</th> <th>Benchmark</th> <th>Qwen/Qwen3.8-27B</th> <th>RedHatAI/Qwen3.8-27B-NVFP4</th> <th>Recovery</th> </tr> </thead> <tbody> <tr> <td rowspan="4"><b>Reasoning</b></td> <td>GSM8K Platinum</td> <td>96.25%</td> <td>96.72%</td> <td>100.49%</td> </tr> <tr> <td>MATH-500</td> <td>83.67%</td> <td>84.27%</td> <td>100.72%</td> </tr> <tr> <td>AIME 2025</td> <td>96.67%</td> <td>95.00%</td> <td>98.27%</td> </tr> <tr> <td>GPQA Diamond</td> <td>89.56%</td> <td>89.22%</td> <td>99.62%</td> </tr> <tr> <td><b>Instruction Following</b></td> <td>IFEval</td> <td>91.19%</td> <td>91.99%</td> <td>100.88%</td> </tr> <tr> <td><b>Agentic - Coding</b></td> <td>SWE Bench</td> <td>78.8%</td> <td>78.0%</td> <td>98.98%</td> </tr> </tbody> </table> NVFP4 build comparison <table> <thead> <tr> <th>Category</th> <th>Benchmark</th> <th>RedHatAI/Qwen3.8-27B-NVFP4</th> <th>unsloth/Qwen3.8-27B-NVFP4</th> <th>Inferact/Qwen3.8-27B-NVFP4</th> </tr> </thead> <tbody> <tr> <td rowspan="4"><b>Reasoning</b></td> <td>GSM8K Platinum</td> <td>96.72%</td> <td>95.42%</td> <td>93.77%</td> </tr> <tr> <td>MATH-500</td> <td>84.27%</td> <td>85.67%</td> <td>82.47%</td> </tr> <tr> <td>AIME 2025</td> <td>95.00%</td> <td>93.75%</td> <td>91.66%</td> </tr> <tr> <td>GPQA Diamond</td> <td>89.22%</td> <td>89.39%</td> <td>87.04%</td> </tr> <tr> <td><b>Instruction Following</b></td> <td>IFEval</td> <td>91.99%</td> <td>91.81%</td> <td>91.50%</td> </tr> </tbody> </table>

Reproduction

The results were obtained using the following commands. Each benchmark was run multiple times with different random seeds — 3 repetitions for GSM8K Platinum, MATH-500, GPQA Diamond, and IFEval, and 8 repetitions for AIME 2025 — and the reported score is the mean across seeds.

<details>

GSM8K Platinum & IFEval (lm-eval, 0-shot)

Run once per seed:

lm_eval --model local-chat-completions \
  --tasks gsm8k_platinum_cot_llama \
  --model_args "model=RedHatAI/Qwen3.8-27B-NVFP4,max_length=69632,base_url=http://127.0.0.1:3235/v1/chat/completions,num_concurrent=32,max_retries=3,tokenized_requests=False,tokenizer_backend=None,timeout=3600" \
  --num_fewshot 0 \
  --apply_chat_template \
  --output_path results_gsm8k_platinum.json \
  --seed 1234 \
  --gen_kwargs "do_sample=True,temperature=1.0,top_p=0.95,top_k=20,max_gen_toks=32000,seed=1234"

lm_eval --model local-chat-completions \
  --tasks ifeval \
  --model_args "model=RedHatAI/Qwen3.8-27B-NVFP4,max_length=69632,base_url=http://127.0.0.1:3235/v1/chat/completions,num_concurrent=32,max_retries=3,tokenized_requests=False,tokenizer_backend=None,timeout=3600" \
  --num_fewshot 0 \
  --apply_chat_template \
  --output_path results_ifeval.json \
  --seed 1234 \
  --gen_kwargs "do_sample=True,temperature=1.0,top_p=0.95,top_k=20,max_gen_toks=32000,seed=1234"

MATH-500, AIME 2025, GPQA Diamond (lighteval, 0-shot)

litellm_config.yaml:

model_parameters:
  provider: hosted_vllm
  model_name: hosted_vllm/RedHatAI/Qwen3.8-27B-NVFP4
  base_url: http://127.0.0.1:3235/v1
  api_key: ''
  timeout: 3600
  concurrent_requests: 32
  generation_parameters:
    temperature: 1.0
    max_new_tokens: 65536
    top_p: 0.95
    top_k: 20
    seed: 1234

Run once per seed (changing seed in the config each time):

lighteval endpoint litellm litellm_config.yaml 'math_500@1@3|0' --output-dir results/ --save-details
lighteval endpoint litellm litellm_config.yaml 'aime25@1@8|0' --output-dir results/ --save-details
lighteval endpoint litellm litellm_config.yaml 'gpqa:diamond@1@3|0' --output-dir results/ --save-details

</details>

Performance Evaluation

Each plot sweeps request load for the math_reasoning and HumanEval benchmark datasets. The x-axis shows per-user interactivity in tokens per second, where higher values mean a snappier response for an individual request. The y-axis shows total server throughput in tokens per second, where higher values mean the system is serving more aggregate load. Each colored line represents a different model and speculator configuration.

The plots demonstrate the benefit of not just quantizing the LLM, but combining it with a trained speculator model, RedHatAI/Qwen3.8-27B-speculator.dspark. Each sweep was done using TP=1,DP=4 on B200s.