Anthropic has released Claude Opus 5, replacing Claude Opus 4.8 as the flagship Opus-tier model while maintaining unchanged pricing of $5 per million input tokens and $25 per million output tokens.
- Thinking is now enabled by default, controlled by the effort parameter, with a breaking change that returns a 400 error if thinking is disabled at xhigh or max effort levels.
- The model ID is claude-opus-5, featuring a 1M-token context window and a maximum output of 128k tokens on the synchronous Messages API.
- On FrontierBench v0.1, Opus 5 scored 43.3% at max effort, outperforming Fable 5 (33.7%) and GPT-5.6 Sol (37.5%).
- Agentic performance saw significant jumps, reaching 70.57% on OSWorld 2.0 and 26.0% on Zapier AutomationBench.
- Opus 5 achieved a verified 30.16% on ARC-AGI-3, roughly four times the previous best score, and scored 56.3% on Humanity's Last Exam without tools.
- Cyber safeguards were adjusted to allow source-code vulnerability finding while blocking exploitation, with prompt injection success rates dropping significantly on Gray Swan benchmarks.
Anthropic positions Opus 5 as approaching the intelligence of Claude Fable 5 at half the price, making it the default model on Claude Max and the strongest option on Claude Pro.