Anthropic Launches Opus 4.7, Weaker Than Mythos Preview

Anthropic has released Claude Opus 4.7, its most powerful publicly available model, touting improvements in coding, image analysis, and instruction-following. However, the company acknowledges it does not advance their capability frontier, as the restricted Mythos Preview model outperformed it on every evaluation. Opus 4.7 serves as a testing ground for new cybersecurity safeguards before a broader Mythos-class release. Anthropic intentionally reduced its cyber capabilities during training and introduced a Cyber Verification Program for security professionals. Pricing remains unchanged at $5 per million input tokens and $25 per million output tokens.