AI2 Reasoning Challenge (Easy set) — grade-school science questions.
Data from LayerLens
As of November 21, 2026, the top-scoring model on ARC Easy is Qwen3 32B at 99.1%, followed by Qwen3 32B at 99.1% and Mistral Medium 3 at 99.1%. 20 models have been evaluated on this benchmark.
Last updated: November 21, 2026
Models
20
Best Score
99.1
Average
97.4
Std Dev
4.4
Provider | Model | Input $/M | Output $/M | ARC Easy | Actions |
|---|---|---|---|---|---|
$0.064 | $0.224 | 99.1 | |||
$0.064 | $0.224 | 99.1 | |||
$0.440 | $2.200 | 99.1 | |||
$2.000 | $8.000 | 99.0 | |||
$0.000 | $0.000 | 98.9 | |||
$0.300 | $2.500 | 98.9 | |||
$0.300 | $2.500 | 98.9 | |||
$0.800 | $3.200 | 98.8 | |||
$2.500 | $10.000 | 98.8 | |||
$0.500 | $2.150 | 98.7 | |||
$0.100 | $0.300 | 98.6 | |||
$0.200 | $0.600 | 98.6 | |||
$0.014 | $0.028 | 98.6 | |||
$0.080 | $0.160 | 98.2 | |||
$0.070 | $0.140 | 97.9 | |||
$0.075 | $0.200 | 97.8 | |||
$0.100 | $0.280 | 97.6 | |||
$0.060 | $0.240 | 97.5 | |||
$0.035 | $0.140 | 95.8 | |||
$0.030 | $0.050 | 78.6 |
Pricing from OpenRouter. Benchmarks from Artificial Analysis.
AI2 Reasoning Challenge (Easy set) — grade-school science questions.
This leaderboard shows all models with ARC Easy benchmark scores, ranked from highest to lowest. Pricing data is included to help you compare performance against cost.
Built by @aellman
Also from the founder: BlobCap, a price index for AI training data deals, showing what AI labs actually pay for data.
2026 68 Ventures, LLC. All rights reserved.