AI Summary

OpenAI updated several evaluation benchmarks for its GPT‑6 Astra model after first announcing the model in a blog post on September 3. The revised scores show Astra performing better on some metrics, while comparable models from rival Anthropic received lower scores. The changes were made during an atypical rollout of the announcement. OpenAI had scheduled the post to go live at 2 p.m. ET, but it was not widely viewable online for almost two additional hours.

AI summaries can be wrong sometimes—always verify important details using the source article.