OpenAI quietly revised Astra evaluation metrics
Fortune reports OpenAI has been updating GPT-6 Astra benchmark scores since launch, with changes that appear to favour the model and an unusual delay in publishing the methodology post. Artificial Analysis subsequently overhauled its Intelligence Index after the dispute drew scrutiny.
- Post-launch metric changes made without announcement, per Fortune sources
- Artificial Analysis drops Astra ranking after overhauling its index methodology
- Follows a pattern of post-launch benchmark disputes across frontier model labs