Full benchmarks coming Q2 2026
Model Security Benchmarks
We run all 100+ attacks against every major model. See which ones actually hold up.
Security Resistance Scores
Q1 2026 ResultsResults locked
GPT-4o
OpenAI
██
resist rate: ██%
Results locked
Claude 3.5 Sonnet
Anthropic
██
resist rate: ██%
Results locked
Gemini 1.5 Pro
Google
██
resist rate: ██%
Results locked
Llama 3.1 70B
Meta
██
resist rate: ██%
Results locked
Mistral Large
Mistral
██
resist rate: ██%
Scores based on 100+ attack vectors across 8 categories. Full methodology published at launch.
Get notified when benchmarks drop
Full benchmarks coming Q2 2026. Be first to see the results.