OpenAI's new GPT-6.2 Astra and Sol models are facing widespread criticism from users across Hacker News, Reddit, and X over several frustrating limitations.
The primary complaints center on premature task termination, where the models stop working on problems before completion. Users also report unnecessary complexity that makes it harder to reach straightforward solutions, and they're annoyed by questions about implementation details they have no way of knowing.
Interface design is another sore spot. Users describe the generated interfaces as limited and repetitively similar across different use cases, suggesting a lack of variety and customization in the models' output capabilities.