Anthropic Launches Claude Opus 5 with Frontier Intelligence at Half the Price of Fable 5

Anthropic Launches Claude Opus 5 with Frontier Intelligence at Half the Price of Fable 5

On July 24, 2026, Anthropic officially released Claude Opus 5, representing a massive upgrade to its Opus frontier tier. The model delivers near-Claude Fable 5 intelligence and industry-leading performance on coding and knowledge-work benchmarks, but at half the price of Fable 5—matching the $5 per million input tokens and $25 per million output tokens pricing of its predecessor, Opus 4.8.

The model represents a step-change improvement across key evaluations:

• Coding and Agency: On SWE-bench Verified, it scores an outstanding 96.0% (compared to Sonnet 5's 85.2% and Opus 4.8's 88.6%). On OSWorld 2.0 (a computer use benchmark), it outperforms every other model at any given cost, surpassing Fable 5's best result at just over a third of the cost.
• Novel Problem Solving: On ARC-AGI 3, which measures novel problem-solving, Opus 5's score is three times as high as the next-best model.
• Alignment and Safety: Anthropic's pre-deployment automated behavioral audit found Opus 5 to be its most aligned model to date, scoring 2.3 on overall misaligned behavior (the lowest of its recent models). It adheres to Claude's Constitution better than Opus 4.8, Sonnet 5, or Fable 5, and remains behind Mythos 5 on dual-use cyber hacking and biology capabilities.

Anthropic also introduced two key platform updates alongside Opus 5:

• Mid-conversation tool changes: Allowing developers to modify Claude's available tools mid-conversation without invalidating the prompt cache.
• Automatic fallbacks: Empowering API requests flagged by safety classifiers on Opus 5 or Fable 5 to automatically route to another model (like Sonnet 5) rather than being blocked.¹