Benchmark
BuzzBench
Scores are percentages.
Scores from EQ-Bench boards.
- Models measured
- 23
- Scores recorded
- 23
- Sources
- 1
- IQ weight
- 2.01
Score distribution
- 48.9%
- Mean
- 51.6%
- Median
- 14.3%
- Std dev
- 31.4%
- 10th percentile
- 67.1%
- 90th percentile
- 9.5%
- Min
This benchmark has a weight of 2.01 for chat models, measured over 23 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro Preview 03-25· EQ-Bench boards | 71.1% |
| 2 | Claude Sonnet 3.7· EQ-Bench boards | 68.2% |
| 3 | GPT 4.1· EQ-Bench boards | 67.1% |
| 4 | DeepSeek R1· EQ-Bench boards | 62.7% |
| 5 | Claude 3.5 Sonnet· EQ-Bench boards | 61.9% |
| 6 | ChatGPT-4o Latest· EQ-Bench boards | 59.5% |
| 7 | o1· EQ-Bench boards | 59.3% |
| 8 | gemini-pro-1.5· EQ-Bench boards | 54.9% |
| 9 | Gemini 2.0 Flash· EQ-Bench boards | 54.4% |
| 10 | Claude Haiku 3.5· EQ-Bench boards | 53.3% |
| 11 | Grok 2· EQ-Bench boards | 51.8% |
| 12 | DeepSeek V3· EQ-Bench boards | 51.6% |
| 13 | Mistral Large Instruct 2411· EQ-Bench boards | 47.5% |
| 14 | O1 Mini· EQ-Bench boards | 46.2% |
| 15 | Llama 3.1 405B Instruct· EQ-Bench boards | 45.9% |
| 16 | Qwen2.5 72B Instruct· EQ-Bench boards | 45.9% |
| 17 | Llama 3.3 70B Instruct· EQ-Bench boards | 45.7% |
| 18 | GPT-3.5 Turbo· EQ-Bench boards | 39% |
| 19 | Phi 4· EQ-Bench boards | 38.7% |
| 20 | Llama 3.1 8B Instruct· EQ-Bench boards | 32.9% |
| 21 | Qwen2.5 7B Instruct· EQ-Bench boards | 31.4% |
| 22 | Llama 3.2 3B· EQ-Bench boards | 26.6% |
| 23 | Llama 3.2 1B Instruct· EQ-Bench boards | 9.5% |
Data credits: EQ-Bench (Samuel J. Paech) — cite paech2023eqbench