TriviaQA
TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents.
The Frontier
Best score over time · one chart, every benchmark
Full rankings
20 models tested · sorted by score
| # | Model | Score |
|---|---|---|
| 1 | 87.5 | |
| 2 | 85.8 | |
| 3 | 84.8 | |
| 4 | 82.9 | |
| 5 | 82.7 | |
| 6 | 82.2 | |
| 7 | 80.0 | |
| 8 | 79.9 | |
| 9 | 79.6 | |
| 10 | 78.9 | |
| 11 | 77.9 | |
| 12 | 75.2 | |
| 13 | 73.9 | |
| 14 | U MPT-30B | 73.6 |
| 15 | 71.9 | |
| 16 | 67.7 | |
| 17 | 64.0 | |
| 18 | 58.1 | |
| 19 | 53.2 | |
| 20 | 45.2 |
Score distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with TriviaQA
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Frequently asked
About TriviaQA
What does TriviaQA measure?
TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents. 20 AI models have been tested on it. Scores range from 45.2 to 87.5 out of 100.
Which model leads on TriviaQA?
Claude 2 from Anthropic leads TriviaQA with a score of 87.5. The median score across 20 tested models is 78.4.
Is TriviaQA saturated?
No · the top score is 87.5 out of 100 (88%). There is still meaningful room for improvement on TriviaQA.
Does TriviaQA predict performance on other benchmarks?
Yes · TriviaQA scores correlate 0.90 with HellaSwag across 16 shared models. Models that do well on TriviaQA tend to do well on HellaSwag.
How often is TriviaQA data refreshed?
BenchGecko pulls updates daily. New model scores on TriviaQA appear as soon as they are published by Epoch AI or the model provider.
- Category
- Knowledge
- Max score
- 100
- Models
- 20
- Updated
- 2024-12-26
Top on TriviaQA
Claude 2 · 87.5GPT-3.5 Turbo (older v0613) · 85.8GPT-4 Turbo · 84.8DeepSeek V3 · 82.9Llama 3.1 405B · 82.7More knowledge benchmarks
Same category · related evaluations