Anthropic has launched Claude Opus 5, a new large language model, which is particularly designed to deliver near-flagship performance at a reduced price.
The Claude AI manufacturer stated the model approaches the capabilities of Claude Fable 5 while costing half as much, making it more suitable for everyday professional use.
Improved performance at reduced cost
Claude Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, matching the pricing of its predecessor, Opus 4.8.
On the contrary, Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens.
The model is now launching across Anthropic's platforms, becoming the default option for Claude Max subscribers and the highest-tier model available on Claude Pro.
Strong benchmark results
Anthropic stated Opus 5 sets the latest records across multiple coding and reasoning benchmarks. On Frontier-Bench v0.1, it scored 43.3%, outperforming both Opus 4.8 (18.7%) and Fable 5 (33.7%).
Additionally, it surpassed Fable 5 on the OSWorld 2.0 computer-use benchmark while operating at roughly one-third of the cost.
The company advised Opus 5 for daily coding, research, and knowledge work, while Fable 5 remains better suited for long-duration autonomous tasks. Users can also adjust an effort dial, enabling them to trade response quality for faster performance and lower token usage.
Enterprise benefits and safety enhancements
Early users reported significant efficiency gains. Legal AI firm Harvey said Opus 5 reduced token usage by 26% while maintaining output quality, while Zapier reported the model successfully completed complex automation tasks that previous Claude models could not.
Anthropic also claims Opus 5 is its most aligned and safest model to date, producing fewer deceptive outputs and showing stronger resistance to manipulation attempts. Although it was not specifically trained for cybersecurity, its performance in vulnerability detection improved substantially.
New features for developers
Alongside Opus 5, Anthropic introduced several developer-focused features, including a Fast Mode that delivers responses about 2.5 times faster, automatic API fallback routing, and support for switching tools mid-conversation without disrupting the prompt cache.
Developers can access the model through the Claude API under the name claude-opus-5.
