Benchmarks / Reasoning
TriviaQA
Uses Epoch AI's external TriviaQA exact-match leaderboard. Score: TriviaQA exact match. Collected by Jake Boggs' benchmark dashboard.
TriviaQA
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Llama 2 70b | none | 87.6% |
| 2 | Llama 6 5b | none | 86.0% |
| 3 | GPT-3.5 | high | 85.8% |
| 4 | GPT-4 | high | 84.8% |
| 5 | Llama 2 34b | – | 84.6% |
| 6 | Llama 3 3b | – | 83.8% |
| 7 | DeepSeek V3 | none | 82.9% |
| 8 | Llama 3.1 405b | none | 82.7% |
| 9 | Mixtral 8x7B Instruct | none | 82.2% |
| 10 | DeepSeek V2 | none | 80.0% |
| 11 | Llama 2 13b | none | 79.6% |
| 12 | Llama 1 3b | – | 77.9% |
| 13 | Mistral | none | 75.2% |
| 14 | Llama 2 7b | none | 73.7% |
| 15 | Qwen 2.5 72B | none | 71.9% |
| 16 | Llama 3 8b | none | 67.7% |