DarkBench — benchmark testing model safety and resistance to adversarial attacks.
Data from LayerLens
As of November 21, 2026, the top-scoring model on DarkBench is Qwen3 32B at 49.2%, followed by Qwen3 32B at 49.2% and Gemini 3 Flash Preview at 49.1%. 7 models have been evaluated on this benchmark.
Last updated: November 21, 2026
Models
7
Best Score
49.2
Average
48.0
Std Dev
2.2
Provider | Model | Input $/M | Output $/M | DarkBench | Actions |
|---|---|---|---|---|---|
$0.064 | $0.224 | 49.2 | |||
$0.064 | $0.224 | 49.2 | |||
$0.500 | $3.000 | 49.1 | |||
$0.500 | $3.000 | 49.1 | |||
$0.550 | $2.200 | 48.6 | |||
$0.100 | $0.300 | 47.7 | |||
$0.800 | $3.200 | 42.7 |
Pricing from OpenRouter. Benchmarks from Artificial Analysis.
DarkBench — benchmark testing model safety and resistance to adversarial attacks.
This leaderboard shows all models with DarkBench 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.