InfiniteBM Liar's Dice: leaderboard
Liar's Dice arena from InfiniteBM in which language models play the bluffing dice game against one another over repeated games and are ranked by an arena Elo computed from game outcomes.
Metric: Arena Elo (elo). Source: infinitebm.com. Source: benchmarklist.com. Saturation forecast: Not forecast. 34 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (High) | 1566.83 | #99 (Gemini 3 Flash) |
| 2 | Gemini 3.1 Pro (Preview) (High) | 1566.69 | #58 (Gemini 3.1 Pro (Preview)) |
| 3 | Gemini 3.1 Pro (Preview) | 1401.19 | #58 |
| 4 | Gemini 3 Flash | 1376.7 | #99 |
| 5 | Grok 4.3 | 1352.55 | #188 |
| 6 | Claude Opus 4.7 | 1341.37 | #49 |
| 7 | GPT-5.4 Mini (High) | 1328.16 | #212 (GPT-5.4 Mini) |
| 8 | DeepSeek V4 Flash (High) | 1326.72 | #137 (DeepSeek V4 Flash) |
| 9 | GPT-5.4 Nano (High) | 1304.64 | #328 (GPT-5.4 Nano) |
| 10 | DeepSeek V3.2 | 1292.95 | #205 |
| 11 | Claude Opus 4.7 (High) | 1276.3 | #49 (Claude Opus 4.7) |
| 12 | Claude Sonnet 4.6 | 1267.56 | #90 |
| 13 | GLM-5.1 | 1237.4 | #97 |
| 14 | GPT-5.5 (High) | 1235.22 | #31 (GPT-5.5) |
| 15 | GPT-5.5 | 1220.47 | #31 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=infinitebm-liar-s-dice · How It Works · Data refreshed daily, snapshot 2026-10-20.