Compare harnesses
Every harness on the bench, side by side on the same 30 tasks.
By model
The full matrix. Cell tint scales with the value; the strongest tint wins the row.
| model |  Claude Code |  Codex |  OpenCode |  Hermes Agent |  Pi Agent |  DeepSeek |  Command Code |  Cline |
|---|---|---|---|---|---|---|---|---|
| DeepSeek V4 ProDeepSeek | 50.0%15/30 | 53.3%16/30 | 50.0%15/30 | 56.7%17/30 | 56.7%17/30 | 53.3%16/30 | 60.0%18/30 | 56.7%17/30 |
| GPT-6 AstraOpenAI | 72.4%21/29 | 75.9%22/29 | 69.0%20/29 | 75.9%22/29 | 69.0%20/29 | — | 75.9%22/29 | 79.3%23/29 |
| Harness | DeepSeek V4 Pro | GPT-6 Astra |
|---|---|---|
| Claude CodeAnthropic | 50.0%15/30 | 72.4%21/29 |
| CodexOpenAI | 53.3%16/30 | 75.9%22/29 |
| OpenCodeSST | 50.0%15/30 | 69.0%20/29 |
| Hermes AgentNous Research | 56.7%17/30 | 75.9%22/29 |
| Pi Agentpi.dev | 56.7%17/30 | 69.0%20/29 |
| DeepSeekDeepSeek | 53.3%16/30 | — |
| Command CodeCommand Code | 60.0%18/30 | 75.9%22/29 |
| ClineCline | 56.7%17/30 | 79.3%23/29 |
Overall
Each column shows that harness's selected results on GPT-6 Astra, across the same 30 tasks. Choose a model to compare results across the suite.
Claude CodeCodexOpenCodeHermes AgentPi AgentCommand CodeCline