reasoning benchmark
TruthfulQA
TruthfulQA is a benchmark to measure whether language models are truthful in generating answers to questions. It comprises 817 questions that span 38 categories, including health, law, finance and politics. The questions are crafted such that some humans would answer falsely due to a false belief or misconception, testing models' ability to avoid generating false answers learned from human texts.
Updated Aug 7, 2026
Published models18
Registry coverage18
MetricScore
EvidenceB
TruthfulQA leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MI | 88.0% | 100.0% | 18 | C | |
| 2 | MI | 77.5% | 94.1% | 18 | C | |
| 3 | IB | 66.9% | 88.2% | 18 | C | |
| 4 | MI | 66.4% | 82.3% | 18 | C | |
| 5 | MI | 64.0% | 76.5% | 18 | C | |
| 6 | NR | 63.3% | 70.6% | 18 | C | |
| 7 | NV | 58.6% | 64.7% | 18 | C | |
| 8 | AC | 58.4% | 58.8% | 18 | C | |
| 9 | AL | 58.3% | 52.9% | 18 | C | |
| 10 | IB | 58.1% | 47.1% | 18 | C | |
| 11 | AC | 57.8% | 41.2% | 18 | C | |
| 12 | CO | 56.3% | 35.3% | 18 | C | |
| 13 | AC | 54.8% | 29.4% | 18 | C | |
| 14 | AC | 54.2% | 23.5% | 18 | C | |
| 15 | AL | 54.1% | 17.6% | 18 | C | |
| 16 | IB | 52.1% | 11.8% | 18 | C | |
| 17 | AC | 50.6% | 5.9% | 18 | C | |
| 18 | MA | 50.3% | 0.0% | 18 | C |
Score distribution
Top published rows on the benchmark's original scale.
TruthfulQA highlights
The top published results on this benchmark's own scale.
What is TruthfulQA?
Definition and scoring fields from the benchmark registry.
TruthfulQA is a benchmark to measure whether language models are truthful in generating answers to questions. It comprises 817 questions that span 38 categories, including health, law, finance and politics. The questions are crafted such that some humans would answer falsely due to a false belief or misconception, testing models' ability to avoid generating false answers learned from human texts.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- TruthfulQA
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- truthfulqa|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about TruthfulQA.
Which model scores highest on TruthfulQA?
MAI-Thinking-1 is currently ranked first with 88.0%.
What does TruthfulQA measure?
TruthfulQA is a benchmark to measure whether language models are truthful in generating answers to questions. It comprises 817 questions that span 38 categories, including health, law, finance and politics. The questions are crafted such that some humans would answer falsely due to a false belief or misconception, testing models' ability to avoid generating false answers learned from human texts.
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.