Anthropic Could Be the Real Winner in OpenAI’s Latest Model Release
Reported by Gizmodo1 source ↓
Early reviewers criticized OpenAI’s GPT-6.2 models for being difficult to steer and less capable than competing models. OpenAI product lead Thibault Sottiaux said the company made changes shortly before release that reviewers may not have tested.
Key points
- Reviewers described GPT-6.2 Astra as frustrating to use and said it often needed more guidance
- OpenAI released GPT-6.2 Astra for complex projects and GPT-6.2 Sol for everyday tasks
- OpenAI product lead Thibault Sottiaux said the models changed in the three or four days before release
Details
OpenAI released GPT-6.2 on Tuesday in two versions. The company positions Astra for complex work such as drafting documents and presentations, while Sol is intended for everyday tasks and to use fewer tokens than GPT-6.1 Sol.
Several early reviewers criticized the models’ usability and performance. Entrepreneur Matt Shumer said they were difficult to use and that Astra repeatedly asked permission for routine steps; Wharton professor Ethan Mollick said Astra could complete an ambitious Blender and Unity project but required more guidance than other models. Developer McKay Wrigley and AI commentator Dan Shipper also posted negative assessments.
OpenAI product lead Thibault Sottiaux said the team made a number of changes in the three or four days before release and suggested early reviewers might not have tested the updated models. Gizmodo reported that OpenAI did not immediately respond to its request for comment.
Written from coverage as of
Images
- Gizmodo
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh