Question Answering on GPQA Diamond (Quality, Cost, and Error Reduction)
100Quality (%)Okshot
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Okshotk=12026.06 | 100 | 0.036 | 100 | |
| SOTA LLM2026.06 | 94 | 0.005 | — | |
| Qwen3-235BCategory=Heuristic2026.06 | 78.8 | 1.452 | — | |
| ENTROROUTER2026.06 | 73.2 | 0.902 | — | |
| FORCCategory=Existing Routing Architectures2026.06 | 71.7 | 0.741 | — | |
| RandomCategory=Heuristic2026.06 | 71.2 | 0.696 | — | |
| Qwen3-30BCategory=Heuristic2026.06 | 69.7 | 0.707 | — | |
| RouteLLMCategory=Existing Routing Architectures2026.06 | 65.2 | 0.271 | — | |
| Qwen3-4BCategory=Heuristic2026.06 | 62.6 | 0.052 | — | |
| Router-R1Category=Existing Routing Architectures2026.06 | 49.5 | 0.001 | — | |
| xRouterCategory=Existing Routing Architectures2026.06 | 42.4 | 0.005 | — |