Claude Opus 4.7 Released: SWE-bench 87.6%, 3x Vision Resolution, Best-in-Class MCP Orchestration

Anthropic·high signal

Anthropic released Claude Opus 4.7 on April 16, with SWE-bench Verified jumping from 80.8% to 87.6% and SWE-bench Pro from 53.4% to 64.3%. Vision resolution increased 3x (up to 2,576px long edge), and MCP-Atlas scored 77.3%, best-in-class for multi-tool orchestration. Pricing stays at $5/$25 per million input/output tokens. Notable regression: BrowseComp dropped from 83.7% to 79.3%, behind GPT-5.4 Pro's 89.3%.