Anthropic 发布 Claude Haiku 5.5，API 价格下调

Anthropic has released Claude Haiku 5.5, its cheapest and most capable small model yet, and its fastest at standard speed. Designed for high-volume, cost-sensitive work, it targets summaries, context compaction, database queries, classification, customer support and browser use. It can also serve as a subagent alongside Sonnet 5.5 or Opus 5.5. An adjustable effort setting, a first for Haiku, lets developers trade cost against intelligence.

For prompts up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens, with cache reads at $0.01 and cache writes at $0.125. Rates quintuple above that threshold. Anthropic says average running costs are around 75% below Haiku 4.5, factoring in the tiers and an updated tokenizer. Around 90% of requests to Haiku 4.5 fell within the cheaper tier.

The model scored 72.4% on the offline subset of OSWorld 2.1, up from 15.7% for Haiku 4.5, and 39.2% on Terminal-Bench 4.0, where its predecessor scored zero. It reached 46.4% on FrontierCode 1.1 and 45.9% on Humanity’s Last Exam without tools, rising to 57.4% with tools. Anthropic still positions Sonnet 5.5 and Opus 5.5 for complex agentic coding, while Haiku is aimed at narrower jobs run repeatedly at scale.

Claude Haiku 5.5 is available now on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure under the model ID claude-haiku-5-5. Its cyber safeguards allow more defensive work than Sonnet 5.5 but still block penetration testing, while biology protections match Sonnet 5, Sonnet 5.5 and Opus 5.

The launch halves Sonnet 5.5 cache-read pricing to $0.10 per million tokens, a change Anthropic says lowers the cost of most agentic work by around 20%. Max 5x subscribers will receive $100 in monthly API credits, Max 20x users $200, and Team subscribers up to $500 pooled across users. Anthropic is also adding beta computer-use and browser-use support to its Python and TypeScript SDKs.

Sources and related context

• Monthly API credits for Max and Team plans: Supports: The help article confirms the $100 and $200 monthly Max credits and the Team credit pool capped at $500, and explains eligibility and how to claim them.
• Browser and computer use with the SDK toolsets: Supports: The documentation confirms beta browser-use and computer-use toolsets in the Python and TypeScript SDKs and explains how developers supply their own browser or desktop drivers.

Source