Weapons of Mass Destruction Proxy — benchmark testing knowledge safety boundaries.
Data from LayerLens
As of November 21, 2026, the top-scoring model on WMDP is Gemini 3 Flash Preview at 86.8%, followed by Gemini 3 Flash Preview at 86.8% and DeepSeek V3 at 71.8%. 8 models have been evaluated on this benchmark.
Last updated: November 21, 2026
Models
8
Best Score
86.8
Average
64.8
Std Dev
23.7
Provider | Model | Input $/M | Output $/M | WMDP | Actions |
|---|---|---|---|---|---|
$0.500 | $3.000 | 86.8 | |||
$0.500 | $3.000 | 86.8 | |||
$0.014 | $0.028 | 71.8 | |||
$2.000 | $8.000 | 71.4 | |||
$2.500 | $10.000 | 69.6 | |||
$0.800 | $3.200 | 67.0 | |||
$0.070 | $0.140 | 58.0 | |||
$0.030 | $0.050 | 6.7 |
Pricing from OpenRouter. Benchmarks from Artificial Analysis.
Weapons of Mass Destruction Proxy — benchmark testing knowledge safety boundaries.
This leaderboard shows all models with WMDP 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.