Anthropic ships Claude Haiku 4.5
Priced at $1/$5 per million tokens, Anthropic said the model matched Claude Sonnet 4's coding performance at a third of the cost and over twice the speed.
- Models & capabilities
- Minor
Anthropic released Claude Haiku 4.5, the latest entry in its cheapest and fastest model tier, priced at $1 per million input tokens and $5 per million output tokens. Anthropic said the model matched the coding performance of Claude Sonnet 4 — its mid-tier model from earlier in 2025 — at roughly a third of the cost and more than double the speed, scoring 73.3% on SWE-bench Verified, a benchmark of real-world software-engineering tasks, and reaching about 90% of the agentic-coding performance of the newer Claude Sonnet 4.5.
Sonnet 4.5 remained Anthropic’s stated best coding model, with Haiku 4.5 pitched as a near-frontier option for latency- and cost-sensitive work. Anthropic suggested the two could be combined, with Sonnet 4.5 orchestrating multiple parallel Haiku 4.5 instances on subtasks. The model shipped immediately across Claude.ai, Claude Code, the API, Amazon Bedrock and Google Cloud Vertex AI.
The release continued a pattern across labs in 2025 of pushing capability down into cheap, fast model tiers rather than only up into ever-larger flagships, on the reasoning that agentic and high-volume workloads are more sensitive to latency and per-token cost than to peak benchmark scores.