Artificial Analysis Coding Agent Performance vs. Task Metrics

Effort sweeps connect in effort order; related model versions connect by family. Bubble area represents time per task, or tokens per task when time is on the x-axis. Data refreshed: 2026-09-29 17:00 UTC

Find a model

Horizontal axis

Vertical axis

Connect model families Pareto line

Data sources · Artificial Analysis Coding Agent Index · updated hourly

Snapshot compiled from Artificial Analysis public comparison tables: Claude Code vs Muse Code · Claude Code vs Codex · Claude Code vs Devin Fusion CLI · Claude Code vs Grok Build · Antigravity SDK vs Codex · Kimi Code CLI vs Muse Code · Claude Code vs Opencode.