claude-fable-5

Claude Fable 5 is Anthropic’s current high-capacity model for the most demanding reasoning and long-term agent tasks. It is suitable for complex research, cross-system workflows, long document synthesis, software engineering, complex planning and tasks requiring continuous status management. The model adopts an adaptive thinking mechanism and allocates reasoning depth according to task difficulty, which is suitable for multiple rounds of tool calls and long task advancement. When using it, clear goals, acceptance criteria, tool permissions, and intermediate product storage methods should be provided to avoid over-expanding the task scope of the model under broad requirements.

Context
1000000 tokens
Max output
128000 tokens
Input
llm image
Output
llm

Pricing

Input: $10 / 1M tokens

Output: $50 / 1M tokens

Cached input: $1 / 1M tokens

Cache write: $20 / 1M tokens

Endpoints

  • anthropic
  • openai

More models

  • vertex/gemini-3.1-flash-lite
  • vertex/gemini-3.8-flash
  • gpt-5.4

    gpt-5.4 是 OpenAI 面向编码和专业工作的 GPT-5 系列模型。官方模型页将其定位为较具成本效益的专业级选择,支持文本与图像输入、文本输出,以及函数调用、结构化输出、网页搜索、文件搜索、代码解释器等工具能力。它适合复杂文档处理、软件开发、业务分析和多步骤任务。

  • glm-5.2

    GLM-5.2 于 2026 年 6 月 17 日正式发布并开源,是 GLM-5 系列的新一代旗舰模型,专为长周期软件工程任务设计。模型延续 MoE 路线,总参数 744B,激活参数约 40B,上下文窗口从 5.1 的 200K 跃升至 100 万 Token,最大输出 131K。关键技术改进包括 IndexShare 机制,在百万上下文下将每 Token 计算量降低 2.9 倍。在 FrontierSWE 基准上,GLM-5.2 得分 74.4,与 Claude Opus 4.8 仅差约 1 分;Terminal-Bench 2.1 从 63.5 跃升至 81.0。模型以 MIT 协议开源,上线首日即完成与华为昇腾、寒武纪等 8 大国产算力平台的适配。