LMGame-Bench 2048 — leaderboard
LLM game-playing benchmark: 2048 puzzle testing numerical strategy and lookahead planning.
Metric: Score. Source: huggingface.co. Status: saturation imminent. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O1 (2024-12-17) | 7580 |
| 2 | O3 (2025-04-16) | 7120 |
| 3 | GPT-5 (High) | 5770.7 |
| 4 | Grok 4 | 4650.7 |
| 5 | O4 Mini (2025-04-16) | 4432 |
| 6 | Grok 3 Mini Beta | 4036 |
| 7 | Gemini 2.5 Pro (Preview 05-06) | 3586.7 |
| 8 | DeepSeek R1 0528 | 3330 |
| 9 | Claude Sonnet 4 (20250514) | 3136 |
| 10 | Claude Opus 4 (20250514) | 3036 |
| 11 | O1 Mini (2024-09-12) | 2757.3 |
| 12 | Claude 3.7 Sonnet (20250219) | 2624 |
| 13 | Qwen 3 235B A22B FP8 | 2144 |
| 14 | GPT-OSS-20B | 2040 |
| 15 | GPT-OSS-120B | 1973.3 |
Interactive version: theaggregate.ai/benchmark?slug=lmgame-bench-2048 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.