Model Overview & Capabilities
Claude Haiku 5.5 is Anthropic’s compact high-speed reasoning model released on October 7, 2026, holding Rank #19 on our global leaderboard with an index score of 38.6. Engineered for rapid agent execution, Haiku 5.5 scores 39.0% on SWE-bench Verified, 38.5% on Terminal-Bench 2.0, 39.0% on Aider Polyglot, and 78.0% on GPQA Diamond graduate scientific evaluations.Anthropic prices Haiku 5.5 at $0.10 per million input tokens and $0.50 per million output tokens, giving an effective blended rate of $0.20 per million tokens under standard 3:1 input:output weighting and directly matching OpenAI’s GPT-6 Luna while delivering an outstanding $0.0052 per index point. Haiku 5.5 processes queries at 210 tokens per second with a swift 0.55-second time to first token.
Supported by a 1,000,000-token context window and prompt caching at $0.01 per million tokens, Claude Haiku 5.5 is exceptionally well-suited for high-throughput subagent loops, rapid test parsing, pull request automated triaging, real-time developer chat assistance, and structured JSON telemetry extraction. Development teams leverage Haiku 5.5 to run millions of continuous integration evaluations daily while preserving cloud infrastructure budgets.
