Anthropic releases Claude Opus 4.7: How to try it, benchmarks, safety

Anthropic has been shipping products and making news at a breakneck pace in 2026, and on Thursday the AI ​​company announced the launch of Claude Opus 4.7.

Claude Opus 4.7 is Anthropic’s smartest model available to the general public. Anthropic notably declared in a press release that Opus 4.7 is not as powerful as Claude Mythos, which Anthropic deemed too dangerous for public release.

Claude Opus is a family of hybrid reasoning models capable of multi-step reasoning and advanced coding. Until the announcement of Claude Mythos on April 7, Claude Opus was considered Anthropic’s most advanced AI model series.

Don’t miss our latest news: add Mashable as a trusted news source in Google.

How to try Claude Opus 4.7

Claude Opus 4.7 is now available through Claude AI, the Claude API, and Anthropic partners such as Microsoft Foundry. The new model has the same price as the Claude Opus 4.6.

However, Anthropic noted that because “Opus 4.7 thinks more at higher levels of effort”, it uses more output tokens than its predecessor. Users can learn more about how to optimize token usage in the Opus 4.7 migration guide.

How Claude Opus 4.7 improves over 4.6

As expected, Claude Opus 4.7 offers improved capabilities across the board.

In particular, Anthropic claims that Claude Opus 4.7 is better at advanced coding tasks, visual intelligence, and document analysis. Anthropic also claims that Opus 4.7 is “sleaker and more creative when performing professional tasks, producing higher quality interfaces, slides and documents.”

“Users report being able to confidently hand over their most difficult coding work – one that previously required close supervision – to Opus 4.7. Opus 4.7 handles complex, long-running tasks with rigor and consistency, pays close attention to instructions, and designs ways to verify its own results before reporting back,” reads an Anthropic blog post.

Claude Opus 4.7: Benchmark performance

Anthropic has published a detailed model sheet describing how Claude Opus 4.7 compares to other Anthropic models and frontier models from OpenAI, Google and xAI.

Opus 4.7 lags behind the previously unreleased Claude Mythos, which Anthropic says scores significantly higher on common benchmarks such as Mankind’s Last Review. “Claude Opus 4.7 performs less well than Claude Mythos Preview on all relevant axes that we measured and does not advance our capability frontier,” the model sheet states. “This means that Claude Opus 4.7 does not prove that AI development has accelerated beyond existing trend lines.

On Humanity’s Last Exam (without tools), Anthropic reports that Claude Opus 4.7 outperforms all other frontier models except Claude Mythos.

• Claude Mythos scored 56.8 percent on HLE

• Claude Opus 4.7 got 46.9 percent

• Gemini 3.1 Pro scored 44.4%

• GPT-5-4 Pro scored 42.7%

• Claude Opus 4.6 scored 40.0 percent

With the tools, GPT-5-4-Pro scored 58.7%, compared to 54.7% for Opus 4.7. Mythos beat them both with 64.7 percent.

Mashable has not independently verified these benchmark results. The complete results are available in the Opus 4.7 model sheet.

Credit: Anthropic

Overall, Anthropic earned a higher Opus 4.7 score than other flagship models in some tests, although Gemini 3.1 Pro and GPT-5-4 scored higher in some areas.

Claude Opus 4.7: Security and hallucinations

Anthropic also reports that Opus 4.7 has a low risk of misaligned behaviors, with a similar risk profile to Opus 4.6.

For example, Anthropic claims that Opus 4.7 is less likely to hallucinate and has lower reward hack rates.

“Claude Opus 4.7 is more reliably honest than Opus 4.6 or Sonnet 4.6, with large reductions in the rate of significant omissions and moderate improvements in factuality and hallucinatory entry rates,” the model card states.

Want to learn more about how to get the most out of your technology? Sign up for Mashable’s Top Stories and Deals newsletters Today.