reasoning benchmark
TriviaQA
A large-scale reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents (six per question on average) that provide high quality distant supervision for answering the questions. The dataset features relatively complex, compositional questions with considerable syntactic and lexical variability, requiring cross-sentence reasoning to find answers.
Updated Aug 7, 2026
Published models18
Registry coverage18
MetricScore
EvidenceB
TriviaQA leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MA | 85.1% | 100.0% | 18 | C | |
| 2 | GO | 83.7% | 94.1% | 18 | C | |
| 3 | XI | 81.3% | 88.2% | 18 | C | |
| 4 | MA | 80.5% | 82.3% | 18 | C | |
| 5 | MA | 80.5% | 76.5% | 18 | C | |
| 6 | MA | 80.3% | 70.6% | 18 | C | |
| 7 | IB | 78.2% | 64.7% | 18 | C | |
| 8 | GO | 76.6% | 58.8% | 18 | C | |
| 9 | MA | 74.9% | 52.9% | 18 | C | |
| 10 | MA | 74.9% | 47.1% | 18 | C | |
| 11 | MA | 73.8% | 41.2% | 18 | C | |
| 12 | GO | 70.2% | 35.3% | 18 | C | |
| 13 | GO | 70.2% | 29.4% | 18 | C | |
| 14 | MA | 68.1% | 23.5% | 18 | C | |
| 15 | MA | 65.5% | 17.6% | 18 | C | |
| 16 | GO | 60.8% | 11.8% | 18 | C | |
| 17 | GO | 60.8% | 5.9% | 18 | C | |
| 18 | MA | 59.2% | 0.0% | 18 | C |
Score distribution
Top published rows on the benchmark's original scale.
TriviaQA highlights
The top published results on this benchmark's own scale.
What is TriviaQA?
Definition and scoring fields from the benchmark registry.
A large-scale reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents (six per question on average) that provide high quality distant supervision for answering the questions. The dataset features relatively complex, compositional questions with considerable syntactic and lexical variability, requiring cross-sentence reasoning to find answers.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- TriviaQA
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- triviaqa|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about TriviaQA.
Which model scores highest on TriviaQA?
Kimi K2 Base is currently ranked first with 85.1%.
What does TriviaQA measure?
A large-scale reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents (six per question on average) that provide high quality distant supervision for answering the questions. The dataset features relatively complex, compositional questions with considerable syntactic and lexical variability, requiring cross-sentence reasoning to find answers.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
18 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.