Coding agents
Compare coding agents and harnesses by how they run and integrate with editors.
| Attribute | AiderAider | Antigravity CLIGoogle |
|---|---|---|
| Product type | Coding harness Aider usage docs · | Coding agent Google Antigravity CLI documentation · |
| Interfaces | Terminal Aider usage docs · | Terminal Google Antigravity CLI documentation · |
| File editing | Yes Aider usage docs · | Yes Google Antigravity CLI documentation · |
| Action approval controls | Unknown | Yes Google Antigravity CLI documentation · |
| Terminal interface | Unknown | Yes Google Antigravity CLI documentation · |
| Individual software fee per month | Unknown | Unknown |
Snapshot · Catalog and evidence ↗
Coding agent benchmark ranking
Agents ranked by their best Terminal-Bench result with any model; the model behind each score is named with it and drives much of the result. Agents documented as unable to edit files are excluded. Agents with no leaderboard submission are listed as not benchmarked.
Terminal-Bench 4.1 · measured 2026-11-19 by Terminal-Bench 4.1 leaderboard. Best public leaderboard resolution rate for each agent with any model. The model behind the score is named with it, and it drives much of the result.
- GitHub Copilot CLI · 71.2% Copilot Fusion · max
- Claude Code · 65.8% Fable 5.2 · max
- Antigravity CLI · 64.5% Gemini 4 Carbon · high
- Codex · 62.1% GPT-6.2 Astra · max
- Muse Code · 47.0% Muse Wonder · xhigh
- Grok Build · 38.8% Grok 4.7 · xhigh file-edits unverified
- mini-swe-agent · 27.0% MiMo V2.6 Pro · none file-edits unverified
Not benchmarked: Aider, Cline, Cursor, OpenCode 2.