Models · 24 November 2025

Claude Opus 4.5 leads on coding benchmarks at a much lower $5/$25 price

Anthropic launched Claude Opus 4.5, reporting state-of-the-art performance on SWE-bench Verified and gains in agentic and computer-use tasks. Pricing fell to $5 per million input tokens and $25 per million output tokens, and Anthropic said the model reaches similar results while using substantially fewer tokens. For practitioners, the price cut made an Opus-class model practical for everyday coding and agent workloads.