#benchmark の記事

Claude Haiku 4.5 Performance | Speed and Cost-Efficiency Verified with Benchmarks

We examine Claude Haiku 4.5's performance across three dimensions: coding accuracy on SWE-bench Verified at 73.3% — on par with the previous-generation Sonnet 4 — response speed 4–5x faster than Sonnet 4.5, and pricing at $1 input / $5 output per million tokens. We also cover the 200K context limit and guidelines for choosing between models.