Qwen3.8-2.4T

Qwen3.8-2.4T-A95B This repository contains 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. For users seeking managed, scalable inference without infrastructure maintenance, the official Qwen API service is provided by Qwen Cloud. Qwen3.8-2.4T-A95B 本仓库包含了 Hugging Face Transformers 格式的后训练模型权重及配置文件。这些模型文件兼容 vLLM、SGLang、TokenSpeed 等框架。对于寻求无需维护基础设施、可扩展托管推理服务的用户，Qwen Cloud 提供了官方 Qwen API 服务。

In particular, Qwen3.8-Max is 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 the Qwen3.8-Max Overview. 特别地，Qwen3.8-Max 是基于 Qwen3.8-2.4T-A95B 构建的官方版本，具备更多功能，例如视觉输入、非思考模式支持、默认 100 万上下文长度、官方内置工具等。更多信息请参考 Qwen3.8-Max 概览。

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. 继 Qwen3.5 和 Qwen3.6 系列在社区获得广泛应用后，我们很高兴推出 Qwen3.8，这是迄今为止 Qwen 开源模型家族中能力最强的版本。Qwen3.8 首次将 Qwen-Max 级别的模型带入开源领域。基于 Qwen3.5 的架构基础，Qwen3.8 在编程、专业工作、研究以及长周期智能体任务方面实现了显著提升。除了回答更具挑战性的问题外，Qwen3.8 还旨在以更高的可靠性完成复杂的多步骤任务。

Qwen3.8 Highlights

Qwen3.8 亮点

Qwen3.8 features the following enhancements: Qwen3.8 具备以下增强功能：

Core Capabilities: Comprehensive improvements across coding, professional work, research, and long-horizon agentic tasks. 核心能力： 在编程、专业工作、研究和长周期智能体任务方面进行了全面改进。
Agent Execution: Stronger autonomous planning and better handling of environment feedback, leading to more reliable end-to-end task completion. 智能体执行： 更强的自主规划能力和对环境反馈的更好处理，从而实现更可靠的端到端任务完成。
Downstream Compatibility: Broader support for popular harnesses and development tools, making it easier to integrate into your existing stack. 下游兼容性： 对主流评测框架和开发工具提供了更广泛的支持，使其更容易集成到您现有的技术栈中。
Flexible Thinking Control: Reasoning depth can be tuned with reasoning_effort, and reasoning context from historical messages is retained via preserve_thinking. 灵活的思考控制： 可以通过 reasoning_effort 调整推理深度，并通过 preserve_thinking 保留历史消息中的推理上下文。

For more details, please refer to our blog post Qwen3.8-Max. 更多详情，请参阅我们的博客文章 Qwen3.8-Max。

Model Overview

模型概览

Type: Causal Language Model 类型： 因果语言模型
Training Stage: Pre-training & Post-training Language Model 训练阶段： 预训练与后训练语言模型
Number of Parameters: 2.4T in total and 95B activated 参数量： 总计 2.4T，激活 95B
Hidden Dimension: 8192 隐藏层维度： 8192
Context Length: 262,144 natively and extensible up to 1,010,000 tokens. 上下文长度： 原生支持 262,144，最高可扩展至 1,010,000 tokens。

(Note: Due to space constraints, the detailed benchmark table and technical footnotes have been omitted from this summary. Please refer to the official repository for the full dataset.) （注：由于篇幅限制，详细的基准测试表格及技术脚注已从本摘要中省略。请查阅官方仓库获取完整数据集。）