Anthropic says an unreleased Claude research model, running a multi-agent swarm in Claude Code, improved a long-standing math bound from 41.6% to 67.2%.
Continue to AI University →