Next-Step-Prediction Style Planning on Cosmos Reason
66.82PerformanceCosmos-Reason2
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Cosmos-Reason2Scale=8B, Mode=Cross-benchmark transfer2026.06 | 66.82 | 53.61 | 6 | |
| Gemini 3 ProScale=-, Mode=Cross-benchmark transfer2026.06 | 64.82 | 58.94 | 1 | |
| Seed2.0 ProScale=-, Mode=Cross-benchmark transfer2026.06 | 63.82 | 57.83 | 2 | |
| Causal PlannerScale=8B, Mode=Cross-benchmark transfer2026.06 | 63.3 | 57.35 | 3 | |
| Cosmos-Reason1Scale=7B, Mode=Cross-benchmark transfer2026.06 | 61.8 | 44.14 | 12 | |
| RoboBrain-2.5Scale=8B, Mode=Cross-benchmark transfer2026.06 | 59.43 | 53.55 | 7 | |
| Qwen3-VLScale=8B, Mode=Cross-benchmark transfer2026.06 | 58.7 | 53.04 | 8 | |
| RynnBrainScale=8B, Mode=Cross-benchmark transfer2026.06 | 57.84 | 54.13 | 4 | |
| Kimi K2.5Scale=1T, Mode=Cross-benchmark transfer2026.06 | 56.82 | 50.26 | 10 | |
| MiMo-EmbodiedScale=7B, Mode=Cross-benchmark transfer2026.06 | 56.8 | 53.93 | 5 | |
| GPT-5.4Scale=-, Mode=Cross-benchmark transfer2026.06 | 55.82 | 52.85 | 9 | |
| Qwen2.5-VLScale=7B, Mode=Cross-benchmark transfer2026.06 | 53.7 | 50.18 | 11 | |
| GPT-4oScale=-, Mode=Cross-benchmark transfer2026.06 | 53.3 | 43.2 | 13 | |
| Gemini 2.5 ProScale=-, Mode=Cross-benchmark transfer2026.06 | 48.64 | 41.8 | 14 | |
| InternVL3.5Scale=8B, Mode=Cross-benchmark transfer2026.06 | 48.24 | 39.9 | 15 | |
| RoboBrain-2.0Scale=7B, Mode=Cross-benchmark transfer2026.06 | 33.82 | 37.79 | 16 |