Anthropic’s Claude Opus 5 resets the agent price/performance benchmark

Large language modelsAgentic AIUnited States

Published: 24 July 2026 (72-hour extension). Opus 5 is available at US$5 per million input tokens and US$25 per million output tokens; Fast mode is advertised at about 2.5 times normal speed for twice the base price. Anthropic positions it for coding, professional work and long-running agents, with new API beta features for mid-conversation tool changes and automatic fallback.

Why it matters: Stronger agent capability at an Opus-tier price can change model-routing economics. Impact: Buyers should benchmark completed-workflow cost, latency, refusal and safety behaviour. Vendor benchmarks and customer testimonials are not independent reproduction. Anthropic · a report