Qwen/Qwen3.8-2.4T-A95B-FP8 · Hugging Face

Source: https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8

This repository contains FP8-quantized model weights and configuration files for the post-trained model in the Hugging Face Transformers format. These artifacts are compatible with vLLM, SGLang, TokenSpeed, etc. The quantization method is fine-grained fp8 quantization with block size of 128, and its performance metrics are nearly identical to those of the original model.

For users seeking managed, scalable inference without infrastructure maintenance, the official Qwen API service is provided byQwen Cloud. In particular,Qwen3.8-Maxis the official version based on Qwen3.8-2.4T-A95B with more features, such as vision input & non-thinking support, 1M context length by default, official built-in tools, etc. For more information, please refer to theQwen3.8-Max Overview.

Following the widespread community adoption of the Qwen3.5 and Qwen3.6 series, we are pleased to introduce Qwen3.8, the most capable generation in the Qwen open-model family to date.

For the first time, Qwen3.8 brings a Qwen-Max-class model to open release. Built on the architectural foundation of Qwen3.5, Qwen3.8 delivers substantial gains across coding, professional work, research, and long-horizon agentic tasks. Beyond answering harder questions, Qwen3.8 is designed to carry complex, multi-step tasks through to completion with greater reliability.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#qwen38-highlightsQwen3.8 Highlights

Qwen3.8 features the following enhancements:

• Core Capabilities: Comprehensive improvements across coding, professional work, research, and long-horizon agentic tasks.
• Agent Execution: Stronger autonomous planning and better handling of environment feedback, leading to more reliable end-to-end task completion.
• Downstream Compatibility: Broader support for popular harnesses and development tools, making it easier to integrate into your existing stack.
• Flexible Thinking Control: Reasoning depth can be tuned withreasoning\_effort, and reasoning context from historical messages is retained viapreserve\_thinking.

For more details, please refer to our blog postQwen3.8-Max.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#model-overviewModel Overview

• Type: Causal Language Model
• Training Stage: Pre-training & Post-training
• Language Model- Number of Parameters: 2.4T in total and 95B activated - Hidden Dimension: 8192 - Token Embedding: 248,320 (Padded) - Number of Layers: 92 - Hidden Layout: 23 × (3 × (Gated DeltaNet → MoE) → 1 × (Gated Attention → MoE)) - Gated DeltaNet:- Number of Linear Attention Heads: 128 for V and 16 for QK - Head Dimension: 128 - Gated Attention:- Number of Attention Heads: 64 for Q and 4 for KV - Head Dimension: 256 - Rotary Position Embedding Dimension: 64 - Mixture of Experts:- Number of Experts: 512 - Number of Activated Experts: 10 Routed + 1 Shared - Expert Intermediate Dimension: 2048 - LM Output: 248,320 (Padded) - MTP (Multi-Token Prediction): trained with multiple steps
• Context Length: 262,144 natively and extensible up to 1,010,000 tokens.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#benchmark-resultsBenchmark Results

Opus 4.8Fable 5GPT 5.6 Sol (max)Qwen3.7-MaxQwen3.8-MaxCoding AgentTerminal Bench 2.184.684.688.874.586.6SWE-bench Pro69.280.064.660.667.7DeepSWE 1.159.070.073.021.656.6NL2Repo-Bench69.4----47.255.9FrontierSWE70.088.8--40.773.5MLS-Bench-Lite42.849.946.231.741.0PaperBench80.388.890.564.893.0AndroidBench69.884.574.056.575.1QwenSWEBench84.086.373.563.480.7QwenQoderBench62.763.153.836.858.4QwenReactBench16941770156415381724QwenSVGBench16481690175814991713General AgentCoWorkBench72.375.971.564.674.8WorkSpaceBench66.868.765.661.467.7JobBench48.457.445.431.353.4SkillsBench65.170.973.561.270.2Agents’ Last Exam (Pass / Score)27.0 / 45.1-- / --30.6 / 53.611.8 / 31.127.0 / 52.4Automation-Bench (Pass@1)27.229.129.714.227.3Toolathlon Verified (Pass@1)76.277.974.949.772.5WideSearch72.981.2--75.281.9HLE w/ tools57.964.558.053.556.2General CapabilitiesGPQA Diamond92.092.694.192.492.6HLE45.753.347.241.443.6IFBench62.263.572.779.182.8$OneMillion-Bench (expert score)41.855.953.844.452.5HealthBench52.4--55.354.560.2PLawBench69.670.272.358.973.2PRBench-Legal52.757.657.648.557.6PRBench-Finance51.955.855.546.858.3MRCR v2 256K (8-needle)83.2--93.886.792.9LongBench v269.1--67.165.366.31. Fable5 results may involve fallbacks. 2. Terminal Bench 2.1: Evaluated with Claude Code (avg@10), using a 5-hour timeout and max_tokens=131,072. For all other models, we report the best published score across harnesses: Claude Opus 4.8 and Claude Fable 5 with Terminus 2 from Artificial Analysis (https://artificialanalysis.ai/evaluations/terminalbench-v2-1); GPT-5.6 Sol with Codex (https://openai.com/index/previewing-gpt-5-6-sol/). 3. SWE-bench Pro: Evaluated with the Claude Code harness, temp=1.0, top_p=0.95, and a 256K context window. Problematic tasks corrected and all baselines evaluated on the refined benchmark. 4. DeepSWE 1.1: Evaluated with the Claude Code and mini-SWE-agent harnesses, temp=1.0, top_p=0.95, and a 256K context window. We report the highest score among both harnesses; notably, Qwen3.8-Max performs best on Claude Code. 5. NL2Repo-Bench: Evaluated with the Claude Code harness. To prevent reward hacking, we disable Bash commands that attempt to access the specific repository, such as pip download, pip install, and git clone. 6. FrontierSWE: Evaluated with the Claude Code harness. All other available MEAN@5 results are taken from the official FrontierSWE leaderboard (https://www.frontierswe.com) as of August 3, 2026. Dominance scores are recomputed from the raw scores using the official evaluation script. “--” indicates that no official MEAN@5 result was available as of that date. 7. MLS-Bench-Lite: Evaluated with Claude Code using a 5-hour timeout and max_tokens=131,072. All other model scores are taken from the official leaderboard. 8. PaperBench: Evaluated in the BasicAgent setting under Code-Dev mode, judged by Claude Opus 4.6, and averaged over 3 runs (max 12 hours per run). 9. AndroidBench: Evaluated on the 95-task public subset, reporting avg@3 scores. 10. QwenSWEBench: Inhouse coding benchmark to evaluate models’ software engineering capabilities. Evaluated with the Claude Code harness. Reporting avg@3 with an 8-hour timeout, max_tokens=32,768, temperature=1.0, and a 256K-token context window. 11. QwenQoderBench: Inhouse coding benchmark to evaluate user experience on Qoder. Evaluated with the Claude Code harness. Reporting avg@5 with a 6-hour timeout, max_tokens=32,768, temperature=1.0, and a 256K-token context window. 12. QwenReactBench: Inhouse React project building benchmark using Claude Code as the harness, bilingual (EN/CN), 7 categories; auto-render + multimodal judge; BT/Elo rating. 13. QwenSVGBench: Inhouse SVG code generation benchmark; bilingual (EN/CN), auto-render + multimodal judge; BT/Elo rating. 14. CoWorkBench: Inhouse cowork benchmark for evaluating long-horizon tasks across computer science, finance, law, medical, and other productivity domains. 15. SkillsBench: Evaluated on the public SkillsBench v1.1 benchmark across 87 tasks, reporting the average score over three runs per task. Opus 4.8 and Fable 5 are evaluated on Claude Code; GPT-5.6 Sol is evaluated on Codex; the Qwen-series are evaluated on OpenCode. All results are from our own testing. 16. Automation-Bench: Evaluated on the 600-task public subset. 17. WideSearch: Evaluated with the Claude Code harness for external models and the Qwen-Agent harness for ours, reporting the average item-F1 over four runs. 18. $OneMillion-Bench: Evaluated using gemini-3.1-pro-preview. 19. PLawBench: Evaluated using gemini-3.1-pro-preview. 20. Empty cells (--): Scores are not yet available or are not applicable.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#quickstartQuickstart

For streamlined integration, we recommend using Qwen3.8 via APIs.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#serving-qwen38Serving Qwen3.8

Inference efficiency and throughput vary significantly across frameworks. We recommend using the latest framework versions to ensure optimal performance and compatibility. For production workloads or high-throughput scenarios, dedicated serving engines such as SGLang, vLLM, or TokenSpeed are recommended.

Qwen3.8 can be deployed with popular inference frameworks, e.g.:

• SGLang:Qwen3.8 Cookbook
• vLLM:Qwen3.8 Recipe
• TokenSpeed:Qwen3.8 Recipe

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#api-usageAPI Usage

Qwen3.8-2.4T-A95B is a text-only model that requires thinking mode for all interactions. Multimodal inputs are not supported, and thinking cannot be disabled. Every response will automatically begin with reasoning enclosed in<think\>\\n\.\.\.</think\>\\n\\nbefore the final output.

We recommend using the following set of sampling parameters for generation: - temperature=1\.0, top\_p=0\.95, top\_k=20, min\_p=0\.0, presence\_penalty=0\.0, repetition\_penalty=1\.0 Please note that the support for sampling parameters varies according to inference frameworks.

Qwen3.8 comes with official support forreasoning\_effort, which can be used to adjust reasoning depth and control cost:

• xhigh(default): for complex tasks demanding thorough analysis
• medium: balancing accuracy and speed
• low: efficient reasoning optimizing for speed and cost

In addition,preserve\_thinkingis enabled by default for all workloads for the best out-of-the-box experience.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#chat-completions-apiChat Completions API

The Chat Completions API can be used with most inference frameworks, as well asQwen Cloud. Before starting, make sure the OpenAI Python SDK is installed and the API key and the API base URL are configured, e.g.:

pip install -U openai

# Set the following accordingly
export OPENAI_BASE_URL='your-base-url'
export OPENAI_API_KEY='your-api-key'

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#text-only-inputText-Only Input

from openai import OpenAI
# Configured by environment variables
client = OpenAI()

messages = [{"role": "user", "content": "Write a Python function to merge two sorted linked lists."}]

completion = client.chat.completions.create(
    model="Qwen/Qwen3.8-2.4T-A95B-FP8",
    messages=messages,
    extra_body={
        "chat_template_kwargs": {
            "enable_thinking": True,  # on by default; should not be turned off
            "preserve_thinking": True, # on by default
        },
    },
    reasoning_effort="xhigh",  # xhigh by default; supported levels are xhigh, medium, and low
    stream=True,
    stream_options={"include_usage": True},
)

reasoning_content = ""
answer_content = ""
is_answering = False
print("\n" + "=" * 20 + "Reasoning" + "=" * 20 + "\n")

for chunk in completion:
    if not chunk.choices:
        print("\nUsage:")
        print(chunk.usage)
        continue

    delta = chunk.choices[0].delta

    if hasattr(delta, "reasoning_content") and delta.reasoning_content is not None:
        if not is_answering:
            print(delta.reasoning_content, end="", flush=True)
        reasoning_content += delta.reasoning_content

    if hasattr(delta, "content") and delta.content:
        if not is_answering:
            print("\n" + "=" * 20 + "Answer" + "=" * 20 + "\n")
            is_answering = True
        print(delta.content, end="", flush=True)
        answer_content += delta.content

If you are using APIs from Qwen Cloud, in addition to changingmodel, please passextra\_body=\{"enable\_thinking": True, "preserve\_thinking": True\}instead ofextra\_body=\{"chat\_template\_kwargs": \{"enable\_thinking": True, "preserve\_thinking": True\}\}.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#best-practicesBest Practices

To achieve optimal performance, we recommend the following settings:

1. Sampling Parameters: - We suggest using the following set of sampling parameters:- temperature=1\.0,top\_p=0\.95,top\_k=20,min\_p=0\.0,presence\_penalty=0\.0,repetition\_penalty=1\.0 - For supported frameworks, you can adjust thepresence\_penaltyparameter between 0 and 2 to reduce endless repetition. However, using a higher value may occasionally result in language mixing and a slight decrease in model performance.
2. Adequate Output Length: To optimize performance on agentic tasks, we recommend allocating sufficient output length to allow the model to generate detailed and comprehensive responses. For frameworks that support separate token limits for internal reasoning and final outputs, we suggest the following configuration within the 1M context length: - **Reasoning Content:**Set the maximum output length to 262,144 tokens. - **Final Response:**Set the maximum output length to 131,072 tokens. These settings provide the necessary capacity for complex reasoning while ensuring ample space for high-quality final deliverables.

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B-FP8#citationCitation

If you find our work helpful, feel free to give us a cite.

@misc{qwen38,
    title = {{Qwen3.8-Max}: A New Bar for Coding and Cowork},
    url = {https://qwen.ai/blog?id=qwen3.8},
    author = {{Qwen Team}},
    month = {August},
    year = {2026}
}