AISchneier on Securityabout 2 hours ago
Anthropic Opus 5 Tops Prompt Injection and ARC-AGI-3 Benchmarks
1 min read
Anthropic's Claude Opus 5 leads the IPI benchmark, cutting attack success within 15 attempts from 5.5% to 2.0%, and tops ARC-AGI-3 with 30.2%, beating GPT-5.6 Sol's previous 7.8% record.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
