| Benchmark | Model | Rank | Results |
|---|---|---|---|
| code-generation-on-terminal-bench-2-1 | HarnessDev Creation (Gemini 3.1 Pro High) (HarnessDev generated harness) | – | Accuracy: 68.8 |
| deep-research-agents-on-browsecomp | HarnessDev Creation (GPT-5.5 creator, Gemini 3.1 Pro executor) | – | Accuracy: 55.4 |