Trending storyPast story
Anthropic breaks down the cost of a single Claude Code task on Opus 5.5
1 report1 sourceupdated 6 days ago
What happened
Summary
Anthropic 发了一篇博客，专门拆解 Opus 5.5 上跑一次 Claude Code 任务的钱花在哪。成本大头是模型推理、工具调用和上下文窗口占用，但正文没披露具体金额或对比数字，更像一份成本透明说明，不是性能报告。所以目前只能知道“哪些环节花钱”，不知道“到底花多少”，这点先别太激动。
Coverage
Follow the reports to see the story from different sides.
Sep 22
- AI HOT (Curated Pool)Anthropic breaks down the cost of a single Claude Code task on Opus 5.5
Anthropic published a blog post that breaks down the cost components of a single Claude Code task on Opus 5.5. The post does not disclose specific dollar amounts or comparisons; it explains that costs come mainly from model inference, tool calls, and context window usage. It reads more as a cost-transparency note than a performance report.