AIFortuneabout 2 hours ago
OpenAI Alters Astra Benchmark Scores Post-Launch
8 min read
OpenAI changed several GPT-6 Astra evaluation metrics after publishing its blog, halving Astra's hallucination rate to 2% before reverting it. Anthropic's Fable 5.1 math score dropped from 87.8% to 78% then rose to 83%. OpenAI attributed changes to evaluation noise and fixes.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
Summary by ByteBrief
