reasoning benchmark
LongFact Concepts
LongFact is a benchmark for evaluating long-form factuality in large language models. It comprises 2,280 fact-seeking prompts spanning 38 topics, designed to test a model's ability to generate accurate, long-form responses. The benchmark uses SAFE (Search-Augmented Factuality Evaluator) to evaluate factual accuracy.
Updated Aug 7, 2026
Published models1
Registry coverage1
MetricScore
EvidenceB
LongFact Concepts leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 0.7% | 100.0% | 1 | C |
LongFact Concepts highlights
The top published results on this benchmark's own scale.
What is LongFact Concepts?
Definition and scoring fields from the benchmark registry.
LongFact is a benchmark for evaluating long-form factuality in large language models. It comprises 2,280 fact-seeking prompts spanning 38 topics, designed to test a model's ability to generate accurate, long-form responses. The benchmark uses SAFE (Search-Augmented Factuality Evaluator) to evaluate factual accuracy.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- LongFact Concepts
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- longfact-concepts|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about LongFact Concepts.
Which model scores highest on LongFact Concepts?
GPT-5 is currently ranked first with 0.7%.
What does LongFact Concepts measure?
LongFact is a benchmark for evaluating long-form factuality in large language models. It comprises 2,280 fact-seeking prompts spanning 38 topics, designed to test a model's ability to generate accurate, long-form responses. The benchmark uses SAFE (Search-Augmented Factuality Evaluator) to evaluate factual accuracy.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
1 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.