Overview
- Anthropic says Haiku 4.5 delivers more than twice the speed of Sonnet 4 at about one‑third the cost for comparable coding tasks.
- API pricing is listed at $1 per million input tokens and $5 per million output tokens, and the model is included for Claude web and app subscribers.
- Company benchmarks report roughly 73% on the SWE‑bench Verified coding test, putting Haiku 4.5 close to Sonnet 4’s score, though these figures are vendor‑provided.
- A described workflow has Sonnet 4.5 decompose complex problems and coordinate many Haiku 4.5 instances in parallel to reduce latency and compute spend.
- The model is available now via the Claude API as well as Amazon Bedrock and Google Cloud Vertex AI, with Anthropic citing improved automated safety results and early testers noting arithmetic mistakes.