Claude Opus 5: The System Card
ai
Anthropic has released Claude Opus 5, a mid-tier model pitched as a practical alternative to its more expensive Fable 5. Per LessWrong analyst Zvi Mowshowitz, Opus 5 matches or exceeds Fable's capabilities on most everyday tasks, runs faster, and costs half as much. On specialized work—like cybersecurity—Opus 5 doesn't reach the level of Anthropic's most powerful model, Mythos 5, but it substantially outpaces its predecessor, Opus 4.8. Anthropic also relaxed some safeguards with this release: Opus 5 now permits vulnerability analysis in source code, a change intended to help developers write more secure software. Safety classifiers trigger less than five percent of the time, a dramatic drop from forty-two percent previously.
Source: https://www.lesswrong.com/posts/ywGX6FhgbZEkHRfQR/claude-...
Listen to this story
Hear this and more stories in a personalized audio briefing.
Open The Chonkerton