Anthropic released Claude Opus 5 on July 24, positioning it as a faster, cheaper successor to Opus 4.8. Pricing holds at $5 per million input tokens and $25 per million output tokens — the same as before — while the company says the model closes much of the performance gap with its most powerful system, Fable 5, at roughly half the practical cost.\n\nOpus 5 doubled Opus 4.8's score on Frontier-Bench, landed within 0.5% of Fable 5 on CursorBench 3.2 at half the cost, and posted a threefold score jump on ARC-AGI 3. Anthropic also introduced an "effort dial" that lets users trade compute for cost on simpler tasks, and highlighted gains in self-verification, scientific research such as organic chemistry and protein analysis, and agentic coding tools like Zapier automation.\n\nThe release intensifies competition among frontier AI labs racing to cut inference costs while improving reasoning and coding performance, a race also visible this week in a flurry of open-weight model releases from Chinese labs.\n\nSources:\n- https://www.anthropic.com/news/claude-opus-5\n- https://techcrunch.com/2026/07/24/anthropic-launches-opus-5/