Benchmarks / Other
Poker Agent
TrueSkill rating for language models playing multiplayer no-limit Texas Hold'em; Vals scales ratings by 40 on a 1000-point baseline.
Poker Agent
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | GPT-5.2 | xhigh | 1132 |
| 2 | GPT-5 | high | 1103 |
| 3 | Gemini 3 Flash | high | 1100 |
| 4 | DeepSeek V3.2 | high | 1090 |
| 5 | Grok 4.1 Fast | high | 1079 |
| 6 | Gemini 3 Pro | high | 1079 |
| 7 | DeepSeek V3.1 | none | 1058 |
| 8 | Sonnet 4.5 | high | 1056 |
| 9 | GPT-5.1 | high | 1039 |
| 10 | Grok 4 Fast | none | 1034 |
| 11 | Opus 4.5 | high | 1033 |
| 12 | Gemini 2.5 Pro | high | 1033 |
| 13 | gpt-oss-120b | high | 1015 |
| 14 | Kimi K2 | medium | 1012 |
| 15 | Qwen 3 Max | max | 995 |
| 16 | GLM 4.6 | none | 946 |
| 17 | Llama 4 Maverick | medium | 891 |