Anthropic Launches Claude Haiku 5.5 at 75% Lower Average Cost

Anthropic released Claude Haiku 5.5 on October 7, 2026, pricing it at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, which Anthropic says is approximately 75% cheaper on average and up to 90% cheaper on shorter requests than its predecessor Claude Haiku 4.5. The model matches OpenAI's GPT-6 Luna on price at that tier, though beyond 100,000 tokens the price jumps fivefold to $0.50/$2.50, making GPT-6 Luna a better deal at that scale according to Simon Willison. Haiku 5.5 requires reasoning and defaults to medium thinking effort; disabling reasoning is not supported. The model also uses a new tokenizer that consumes roughly 1.25x more tokens for the same prompt compared to Haiku 4.5, which effectively raises costs beyond the nominal rate. Anthropic describes it as the cheapest, fastest, and most capable small model it has released. On the OSWorld 2.1 computer-use agent benchmark, Haiku 5.5 scored 72.4% compared to 15.7% for Haiku 4.5; on the Chartography visual chart-reading benchmark, it scored 46.4% compared to 6.4%. Haiku 5.5 introduces the first effort setting for the Haiku line, allowing users to trade cost for accuracy, though Anthropic recommends Sonnet 5.5 and Opus 5.5 for complex agentic coding tasks. Asana reported over 30% lower task-completion latency and up to 2.5x faster inference per agent turn using the model. The release also included a halving of Sonnet 5.5 cache-read prices, and paid plans now include monthly API credits of $100 for Max 5x, $200 for Max 20x, and up to $500 pooled for Team.