Anthropic has released Claude Opus 4.7, a model intentionally built to be less capable than the upcoming Claude Mythos. While Opus 4.7 brings improvements in software engineering, vision (3x higher resolution support up to 2,576px), memory, instruction-following, and financial analysis, Anthropic openly acknowledges it is 'less broadly capable' than Mythos Preview. The deliberate capability reduction — especially in cybersecurity — is a safety strategy: Anthropic is using Opus 4.7 as a testbed for cyber safeguards before applying lessons to the more powerful Mythos. Mythos Preview outperforms Opus 4.7 on several benchmarks including SWE-Bench Pro and Humanity's Last Exam. Opus 4.7 is available via Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry at the same pricing as Opus 4.6 ($5/$25 per million tokens).
Sort: