qwen3.7-max

Qwen3.7-Max was released on May 20, 2026. It is the flagship version of Qianwen 3.7 series and is positioned as "a new generation flagship model for the era of intelligent agents." The model has more than 1 trillion parameters and 1 million Token context windows. The core advantage lies in the breadth and depth of the agent's capabilities, with outstanding performance in programming, office productivity, and long-term autonomous execution. It scored 72.3 on Terminal-Bench 2.0 Terminus and 60.4 on SWE-Pro. Many indicators refreshed the domestic SOTA. Its signature capability is to support 35 hours of continuous autonomous operation, complete more than 1,100 tool calls, and achieve a 10.0x geometric mean speedup in fully autonomous kernel optimization tasks. The model is compatible with mainstream agent frameworks such as Claude Code and OpenClaw, and is a reliable base for building complex Agent systems.

Context
1000000 tokens
Max output
131072 tokens
Input
llm
Output
llm

Pricing

Input: $2.5 / 1M tokens

Output: $7.5 / 1M tokens

Cached input: $0.5 / 1M tokens

Cache write: $3.125 / 1M tokens

Endpoints

  • openai

More models

  • grok-4.3

    Grok 4.3 于2026年5月发布,是 xAI 在 xAI 并入 SpaceX 更名 SpaceXAI 之前推出的推理模型,定位为面向高级逻辑、数学、科学分析和多步调查等复杂任务的通用旗舰。相比前代 Grok 4.20,Grok 4.3 的架构经过改进,上下文窗口扩展至100万 Token,知识截止日期为2025年12月。模型支持文本和图像输入、文本输出,具备函数调用、结构化输出和缓存输入 Token 能力。

  • gpt-5.3-codex

    gpt-5.3-codex 是 OpenAI 面向代理式软件工程的 Codex 模型标识。官方模型目录将 Codex 系列用于长链路编码任务,例如理解代码库、规划修改、编辑多文件、运行测试、排查错误和调用开发工具。它适合在明确仓库范围、测试标准和变更约束下协助研发工作。调用时应提供项目结构、语言版本、依赖、允许访问的工具和不可修改区域,并将生成的变更置于版本控制、自动化测试和安全审查之下。

  • glm-5.3
  • gpt-5.6-luna

    gpt-5.6-luna 是 OpenAI GPT-5.6 系列中面向成本敏感、高并发场景的模型。官方资料将其定位为较低成本的工作负载选择,同时保留文本与图像输入、文本输出和推理能力,可用于批量信息抽取、分类、摘要、知识库问答和轻量代理步骤。它更适合作为分层架构中的高频执行模型,而复杂推理或关键决策可交由更高层级模型处理;应通过业务样本评估其准确性与重试策略。