Leaderboard: Top OSS Coding models
1.

Kimi K2
2.

DeepSeek V3
3.

Qwen 2.5 Coder 32B
4.

Qwen 2.5 72B
5.

Nemotron 70B
6.

Llama 3.3 70B
7.

Llama 3.1 405B
8.

Llama 3.1 70B
14.

Mistral Small 3
| Model | Organization | Total games | Win % | Playground |
---|
1. | Kimi K2 | Moonshot | 127 | 72% | |
2. | DeepSeek V3 | DeepSeek | 81 | 68% | |
3. | Qwen 2.5 Coder 32B | Qwen | 835 | 54% | |
4. | Qwen 2.5 72B | Qwen | 777 | 50% | |
5. | Nemotron 70B | NVIDIA | 785 | 50% | |
6. | Llama 3.3 70B | Meta | 1261 | 50% | |
7. | Llama 3.1 405B | Meta | 1034 | 49% | |
8. | Llama 3.1 70B | Meta | 955 | 47% | |
14. | Mistral Small 3 | Mistral AI | 76 | 20% | |