Agent Evaluation on OpenThoughts TBLite
72.1AccuracyKimi-K2.5
Evaluation Results
| Method | Links | |
|---|---|---|
| Kimi-K2.5Scale Band=Frontier, Agent Harness=terminus-22026.06 | 72.1 | |
| OpenThinkerAgent-32BScale Band=32B, Agent Harness=terminus-22026.06 | 41.3 | |
| Nemotron-Terminal-8BScale Band=8B, Agent Harness=terminus-22026.06 | 25.8 | |
| OpenThinker-Agent-v1Scale Band=8B, Agent Harness=terminus-22026.06 | 10.9 | |
| SWE-Lego-Qwen3-8BScale Band=8B, Agent Harness=OpenHands2026.06 | 3.2 |