AI2 Reasoning Challenge (Challenge set) — grade-school science questions requiring complex reasoning.
Data from LayerLens
As of November 21, 2026, the top-scoring model on ARC Challenge is GPT-5 at 96.3%, followed by GLM 5 at 96.0% and GLM 5 at 96.0%. 22 models have been evaluated on this benchmark.
Last updated: November 21, 2026
Models
22
Best Score
96.3
Average
92.2
Std Dev
8.1
Provider | Model | Input $/M | Output $/M | ARC Challenge | Actions |
|---|---|---|---|---|---|
$1.250 | $10.000 | 96.3 | |||
$0.600 | $1.920 | 96.0 | |||
$0.600 | $1.920 | 96.0 | |||
$0.014 | $0.029 | 95.4 | |||
$0.000 | $0.000 | 95.3 | |||
$2.000 | $8.000 | 95.1 | |||
$0.500 | $2.150 | 95.1 | |||
$0.200 | $0.600 | 95.0 | |||
$0.000 | $0.000 | 94.8 | |||
$0.064 | $0.224 | 94.7 | |||
$0.064 | $0.224 | 94.7 | |||
$0.014 | $0.028 | 94.0 | |||
$0.100 | $0.300 | 93.9 | |||
$0.440 | $2.200 | 93.9 | |||
$2.500 | $10.000 | 93.7 | |||
$0.440 | $2.200 | 93.4 | |||
$0.800 | $3.200 | 92.7 | |||
$0.070 | $0.140 | 92.6 | |||
$0.075 | $0.200 | 91.8 | |||
$0.080 | $0.160 | 91.0 | |||
$0.035 | $0.140 | 86.9 | |||
$0.030 | $0.050 | 56.3 |
Pricing from OpenRouter. Benchmarks from Artificial Analysis.
AI2 Reasoning Challenge (Challenge set) — grade-school science questions requiring complex reasoning.
This leaderboard shows all models with ARC Challenge 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.