Agentic Coding on TerminalBench 2
81.8Pass RateForgeCode
Evaluation Results
| Method | Links | |
|---|---|---|
| ForgeCodeBase Model=Claude Opus 4.6, Auto=false2026.03 | 81.8 | |
| Meta-HarnessBase Model=Claude Opus 4.6, Auto=true2026.03 | 76.4 | |
| CapyBase Model=Claude Opus 4.6, Auto=false2026.03 | 75.3 | |
| Terminus-KIRABase Model=Claude Opus 4.6, Auto=false2026.03 | 74.7 | |
| MAYA-V2Base Model=Claude Opus 4.6, Auto=false2026.03 | 72.1 | |
| TongAgentsBase Model=Claude Opus 4.6, Auto=false2026.03 | 71.9 | |
| DroidBase Model=Claude Opus 4.6, Auto=false2026.03 | 69.9 | |
| MuxBase Model=Claude Opus 4.6, Auto=false2026.03 | 66.5 | |
| Terminus 2Base Model=Claude Opus 4.6, Auto=false2026.03 | 62.9 | |
| Claude CodeBase Model=Claude Opus 4.6, Auto=false2026.03 | 58 | |
| Meta-HarnessBase Model=Claude Haiku 4.5, Auto=true2026.03 | 37.6 | |
| GooseBase Model=Claude Haiku 4.5, Auto=false2026.03 | 35.5 | |
| Terminus-KIRABase Model=Claude Haiku 4.5, Auto=false2026.03 | 33.7 | |
| Mini-SWE-AgentBase Model=Claude Haiku 4.5, Auto=false2026.03 | 29.8 | |
| Terminus 2Base Model=Claude Haiku 4.5, Auto=false2026.03 | 28.3 | |
| Claude CodeBase Model=Claude Haiku 4.5, Auto=false2026.03 | 27.5 | |
| OpenHandsBase Model=Claude Haiku 4.5, Auto=false2026.03 | 13.9 |