Anthropic Releases Claude Opus 4.7, its Most Intelligent Publicly Available Model

Anthropic announced Claude Opus 4.7 on Thursday, positioning it as the company's most intelligent model available to the general public—though notably less capable than Claude Mythos, which Anthropic deemed too dangerous for release. Opus 4.7 is a hybrid reasoning model that excels at multi-step reasoning and advanced coding. According to Anthropic, users can now hand off their hardest coding work—tasks that previously required close supervision—with confidence, as the model handles complex, long-running tasks with rigor and consistency, pays precise attention to instructions, and even devises ways to verify its own outputs before reporting back.

On benchmarks, Opus 4.7 outperforms all other frontier models except Mythos on Humanity's Last Exam without tools, scoring 46.9% versus GPT-5-4 Pro's 42.7% and Gemini 3.1 Pro's 44.4%. With tools, GPT-5-4 Pro leads at 58.7% compared to Opus 4.7's 54.7%. The model is priced the same as Opus 4.6 but uses more output tokens at higher effort levels. Anthropic also reports reduced hallucination rates and lower reward hacking compared to Opus 4.6. Available now via Claude AI, API, and partners like Microsoft Foundry.