Anthropic has unveiled Claude Opus 5, a model whose capabilities “approach the intelligence” of the flagship Fable 5. The company has made the model immediately available across all platforms.
It has become the default model in the Max subscription, with prices remaining at the Opus 4.8 level—$5 per million input tokens and $25 per million output tokens.
Anthropic claims that Opus 5 has reached the level of Claude Fable 5 at half the token cost. The company also stated that the model has “set a new benchmark” in tests evaluating programming and intellectual skills, such as Frontier-Bench and GDPval-AA, although it still lags behind Mythos 5 in cybersecurity tasks.
According to Anthropic, in the Frontier-Bench v0.1 benchmark, Opus 5 more than doubled the performance of Opus 4.8 at a lower cost per task. In CursorBench 3.2, the gap from the peak result of Fable 5 was less than 0.5% at half the cost.
In the ARC-AGI 3 test, Opus 5 achieved results three times higher than its nearest competitor. On the Zapier AutomationBench platform, which evaluates business task solutions, the success rate was about 1.5 times higher at equal financial costs. In the OSWorld 2.0 test (computer performance evaluation), the model surpassed the best result of Fable 5.
Anthropic highlighted progress in scientific tasks. According to the company, Opus 5 outperformed Opus 4.8 in all life sciences assessments. In the internal organic chemistry benchmark, the result increased by 10.2 percentage points, and in protein-related tasks, by 7.7 percentage points.
According to Anthropic, Opus 5 has become the company’s most compliant model. In an automated behavioral audit, the overall non-compliance score was 2.3, the best among Anthropic’s latest models.
For cybersecurity tasks, Anthropic has relaxed restrictions compared to Fable 5. The company estimated that classifier interventions would trigger about 85% less frequently. However, requests flagged by security systems in Claude.ai, Claude Code, and Claude Cowork will be redirected to Opus 4.8 by default. A similar switch can be enabled in the API.
Anthropic also noted that biological requests blocked on Fable 5 will now be directed to Opus 5 instead of Opus 4.8.
Along with the model, Anthropic launched two updates in beta: the ability to change available tools during a dialogue without resetting the prompt cache and automatic switches to backup models in the API.
Previously, Anthropic introduced Opus 4.8 in May, simultaneously adding a dynamic workflow feature to Claude Code.
