Anthropic ships Claude Opus 4.7

Anthropic has started publishing the developer paperwork for Claude Opus 4.7, telling customers to expect behavior changes and some breaking changes when they switch. (platform.claude.com) The company’s “What’s new” page says Opus 4.7 is its most capable generally available model and flags changes in model behavior, tool use, and API handling that developers need to review before production rollouts. (platform.claude.com) Anthropic also published a migration guide on April 23 for teams moving from earlier Claude versions to Claude Opus 4.7 and Claude 4.6, with separate instructions for Messages API users and Claude Managed Agents users. (platform.claude.com) A large language model is the text-and-code engine behind a chatbot, and a migration guide is the checklist developers use when a new model may answer differently, call tools differently, or reject old prompts. Anthropic’s docs say Opus 4.7 should work with many existing Opus 4.6 prompts at the same listed price, but they still call out a “handful” of behavior and API changes. (platform.claude.com) Anthropic introduced Claude Opus 4.7 publicly on April 16, saying the model was generally available and stronger than Opus 4.6 on advanced software engineering and other long-horizon tasks. Anthropic’s model overview lists Opus 4.7 at $5 per million input tokens and $25 per million output tokens, with a 1 million-token context window and 128,000-token max output. (anthropic.com) (platform.claude.com) The documentation push lands during a messy week for Anthropic’s developer-facing products. Ars Technica reported the company briefly tested removing Claude Code from some new $20 Pro subscriptions, and Anthropic said the experiment affected about 2% of new prosumer signups rather than existing Pro and Max users. (arstechnica.com) At the same time, Anthropic said it was investigating a report that unauthorized users accessed Claude Mythos Preview through a third-party vendor environment. The Guardian reported Mythos is a restricted cybersecurity model that Anthropic has said could help find dangerous software vulnerabilities. (theguardian.com) Anthropic’s own release notes add another deadline for developers: the company said on April 14 that the Claude Sonnet 4 and Claude Opus 4 API models are deprecated, with retirement scheduled for June 15, 2026, and it recommends moving to Claude Sonnet 4.6 and Claude Opus 4.7. (platform.claude.com) For developers, the immediate task is less about trying a new chatbot than checking prompts, tool calls, and model names before June. Anthropic has now made that migration path explicit, even as questions around pricing, access, and control of its most powerful systems keep piling up. (platform.claude.com 1) (platform.claude.com 2)