Back to BenchmarksFrontier
SimpleBench
Simple reasoning tasks designed to test fundamental capabilities
Leaderboard
(27 models)| Rank | Model | Score | Stderr |
|---|---|---|---|
| 1 | Gemini 3 Pro | 76.40 | — |
| 2 | Gemini 2.5 Pro (Jun 2025) | 62.40 | — |
| 3 | Claude Opus 4.5 | 62.00 | — |
| 4 | GPT-5.2 | 61.60 | — |
| 5 | Grok 4 | 60.50 | — |
| 6 | o3 | 53.10 | — |
| 7 | Claude 3.7 Sonnet | 46.40 | — |
| 8 | o1 | 41.70 | — |
| 9 | DeepSeek V3 | 40.80 | — |
| 10 | o4-mini (high) | 38.70 | — |
| 11 | Grok-3 mini | 36.10 | — |
| 12 | GPT-4.1 | 34.50 | — |
| 13 | Qwen3-235B-A22B | 31.00 | — |
| 14 | DeepSeek R1 | 30.90 | — |
| 15 | Gemini 2.0 Flash Thinking Exp | 30.70 | — |
| 16 | Llama-4-Maverick-17B-128E-Instruct | 27.70 | — |
| 17 | Gemini 1.5 Flash | 27.10 | — |
| 18 | kimi-k2-thinking (official) | 26.30 | — |
| 19 | GPT-4 Turbo | 25.10 | — |
| 20 | Claude 3 Opus | 23.50 | — |
| 21 | Llama 3.1 405B | 23.00 | — |
| 22 | Mistral Large | 22.50 | — |
| 23 | GPT-OSS 120B | 22.10 | — |
| 24 | Llama 3.3 70B | 19.90 | — |
| 25 | GPT-4o | 17.80 | — |
| 26 | c4ai-command-a-03-2025 | 17.40 | — |
| 27 | gpt-4o-mini-2024-07-18 | 10.70 | — |
Data source: Epoch AI, “Data on AI Benchmarking”. Published at epoch.ai
Licensed under CC-BY 4.0