reasoning benchmark
ARC-C
The AI2 Reasoning Challenge (ARC) Challenge Set is a multiple-choice question-answering benchmark containing grade-school level science questions that require advanced reasoning capabilities. ARC-C specifically contains questions that were answered incorrectly by both retrieval-based and word co-occurrence algorithms, making it a particularly challenging subset designed to test commonsense reasoning abilities in AI systems.
Updated Aug 7, 2026
Published models34
Registry coverage34
MetricScore
EvidenceB
ARC-C leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
Score distribution
Top published rows on the benchmark's original scale.
ARC-C highlights
The top published results on this benchmark's own scale.
What is ARC-C?
Definition and scoring fields from the benchmark registry.
The AI2 Reasoning Challenge (ARC) Challenge Set is a multiple-choice question-answering benchmark containing grade-school level science questions that require advanced reasoning capabilities. ARC-C specifically contains questions that were answered incorrectly by both retrieval-based and word co-occurrence algorithms, making it a particularly challenging subset designed to test commonsense reasoning abilities in AI systems.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- ARC-C
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- arc-c|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about ARC-C.
Which model scores highest on ARC-C?
MiMo-V2.5-Pro is currently ranked first with 97.2%.
What does ARC-C measure?
The AI2 Reasoning Challenge (ARC) Challenge Set is a multiple-choice question-answering benchmark containing grade-school level science questions that require advanced reasoning capabilities. ARC-C specifically contains questions that were answered incorrectly by both retrieval-based and word co-occurrence algorithms, making it a particularly challenging subset designed to test commonsense reasoning abilities in AI systems.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
34 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.