The Story

Anthropic has released Claude Opus 4.7, its latest publicly available flagship model focused on improved reasoning and coding performance. As per Anthropic's official announcement, the update builds on Opus 4.6 with measurable gains in developer-focused benchmarks, positioning it as one of the strongest models currently available to the public.

The Details

• Benchmark improvement: Opus 4.7 improves SWE-bench Pro coding performance from 53.4% in 4.6 to 64.3%, marking a significant step forward.
• Still behind internal frontier: Anthropic's unreleased Mythos Preview reportedly achieves higher scores, highlighting a gap between public and internal models.
• Same pricing, higher usage: The model is priced similarly to Opus 4.6, though it may consume tokens faster depending on usage patterns.
• Developer tooling updates: Improvements include updated Claude Code defaults and new commands such as /ultrareview for identifying bugs and design issues.
• Mixed early feedback: While benchmarks show gains, some users have reported varied real-world performance compared to expectations.

Why it matters

Anthropic is increasingly operating on a dual-track strategy: rapid public model releases alongside more powerful, tightly controlled frontier systems like Mythos. This approach allows the company to iterate quickly while reserving its most advanced capabilities for restricted environments. It also signals a growing trend in the industry where publicly available models may no longer fully represent the true cutting edge of AI development.