RSI Arena Stage 1 progress
Agents · 2026-10-24
The median, over 12 benchmarks, of how far an agent's trained model moved from the base model toward NVIDIA's own post-trained model. A value of 0 means no better than the base model, and 100 means level with NVIDIA's model.
Benchmark source: Rsiarena
Top models (higher is better)
| Model | Score | Source |
|---|---|---|
| GPT-6 Astra | 76.0 | RSI Arena paper Table 3 (GPT-6 Astra, progress) · Bake AI |
| MiniMax M3.1 Flash | 57.0 | RSI Arena paper Table 3 (MiniMax M3.1 Flash, progress) · Bake AI |
| GLM-5.3 | 52.0 | RSI Arena paper Table 3 (GLM-5.3, progress) · Bake AI |
| Grok 4.7 | 47.0 | RSI Arena paper Table 3 (Grok 4.7, progress) · Bake AI |
| Muse Spark 1.3 | 45.0 | RSI Arena paper Table 3 (Muse Spark 1.3, progress) · Bake AI |
| MiMo-V2.6-Pro | 42.0 | RSI Arena paper Table 3 (MiMo-V2.6-Pro, progress) · Bake AI |
| DeepSeek V4.1 Flash | 41.0 | RSI Arena paper Table 3 (DeepSeek V4.1 Flash, progress) · Bake AI |
| Gemini 3.8 Flash | 41.0 | RSI Arena paper Table 3 (Gemini 3.8 Flash, progress) · Bake AI |
| Kimi K3 | 35.0 | RSI Arena paper Table 3 (Kimi K3, progress) · Bake AI |
| Step 5 Preview | 32.0 | RSI Arena paper Table 3 (Step 5 Preview, progress) · Bake AI |