Anthropic's Claude Opus 4.7 boosts coding benchmarks by 10%, adds multi-agent coordination

Anthropic's Claude Opus 4.7, released April 16, is the company's most powerful production model designed for complex reasoning, autonomous task execution, and multi-agent coordination. It achieves 87.6% on SWE-bench Verified (up from 80.8% on Opus 4.6), 64.3% on SWE-bench Pro (up from 51.9%), and leads MCP-Atlas with 77.3%. The model features a 1M-token context window with improved long-context retrieval, enabling accurate handling of massive codebases and long conversations. It also introduces self-verification of outputs before returning results, improving reliability in production pipelines.

Vision capabilities have been upgraded to 3.75 megapixels (over 3x previous models), allowing detailed interpretation of screenshots, diagrams, and mockups. Opus 4.7 supports file-system-based memory for persistence across sessions, and strict literal instruction following for predictable behavior. Early testers report delegating complex coding work that previously required continuous supervision. The model also orchestrate other AI agents in parallel, completing tasks faster. It is available now via Overchat AI. Anthropic positions Opus 4.7 below the experimental Mythos Preview but notes it is the first production model to test real-world cybersecurity safeguards.