Anthropic has launched Claude Opus 5, a new model designed for coding, knowledge work, and agent workflows, offering performance improvements and increased efficiency over the previous version.

Anthropic has released Claude Opus 5, the latest iteration of its AI model, following the earlier releases of Sonnet 5, Mythos 5, and Fable 5. The new model is designed to enhance coding capabilities, knowledge work, and agent-based workflows. Anthropic states that Opus 5 provides performance near that of the Fable 5 model while operating at half the price.

Performance and Efficiency Gains

According to Anthropic, Opus 5 achieves similar performance to previous models while generating 26% fewer tokens on average at max reasoning. This reduction in token usage is intended to allow for more complex tasks within a single session. The model also features improved autonomy, better error recovery, and more frequent self-checks during tasks.

New Features and Developer Tools

Alongside the release, Anthropic introduced several new features, including a research preview of 'Fast mode' for the Claude platform, which offers 150% faster token generation at double the standard price. Other updates include automatic fallbacks, where the API switches to another model if a request is declined by a safety classifier, and the ability for developers to swap external tools mid-conversation without clearing the prompt cache.

Pricing and Availability

Claude Opus 5 is available now to all user tiers that previously had access to Opus 4.8. Pricing remains the same as the previous version, with token API costs set at $5 per million input tokens and $25 per million output tokens. Anthropic continues to recommend the Fable 5 model for highly ambitious, long-term autonomous projects.