Anthropic launches Claude Opus 5 as a cost-efficient daily default near the frontier

[Current Developments] AI20 editorial coverage based on Anthropic (2026-07-24). On 24 July 2026, Anthropic introduced Claude Opus 5, describing it as available the same day and positioned as a thoughtful, proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price. Anthropic says Opus 5 is the new default on Claude Max and the strongest model on Claude Pro, with API pricing at $5 per million input tokens and $25 per million output tokens — the same as Opus 4.8 — plus a Fast mode at roughly 2.5× default speed. The company reports state-of-the-art results on coding and knowledge-work evaluations such as Frontier-Bench and GDPval-AA while remaining behind Mythos 5 on cybersecurity tasks, and emphasises stronger verification, iteration and agentic thoroughness in early-access testing. Anthropic also details alignment and safety findings, including an automated behavioural audit score it presents as its most aligned recent model, and safeguard designs that allow beneficial cyber and biology uses while applying stronger guardrails on a narrow range of cyber tasks. Developers can call claude-opus-5 on the Claude API. Why it matters: Price–performance shifts change which models become everyday enterprise defaults. Buyers should refresh evaluation harnesses, safety assumptions and fallback policies when a new Opus-class model is positioned as the daily workhorse. AI20 perspective: Frontier access without shared evaluation and safeguard literacy fragments trust. AI20 should help members compare capability claims against governance evidence — system cards, cyber/biology limits and deployment defaults — not only leaderboard deltas. (Category: Research; Region: Global; Tags: Anthropic; Claude Opus 5; frontier models; coding; safety; API. Original source: https://www.anthropic.com/news/claude-opus-5)