# Here’s Why Some Say OpenAI’s Latest Model Is ‘Unusable,’ ‘Uneven’ And ‘Insufferable To Work With’

- URL: https://talkfuse.com/story/heres-why-some-say-openais-latest-model-is-unusable-uneven-and-insufferable-to-work-with/
- Published: 2026-11-22T15:00:51+00:00
- Updated: 2026-11-22T15:41:23+00:00
- Original publisher: The Daily Caller (publisher lean: Right)
- Original article: https://dailycaller.com/2026/11/22/openai-chatgpt-astra-model-critics-usability
- Topics: AI, Politics, Technology

## Key points

- Developers report that GPT-6.2 Astra is frustrating and hard to steer.
- Testers say it can still handle some ambitious projects well.
- OpenAI says it will prioritise making future models easier to use.

## Summary

OpenAI launched GPT-6.2 Astra, a frontier model aimed at developers tackling demanding projects, alongside its general-use model GPT-6.2 Sol. Some developers and researchers say Astra is difficult to guide, repetitive or frustrating to use, raising questions about whether it is a practical improvement despite its reported capabilities.

Comments from testers describe mixed results: Ethan Mollick said Astra could complete an ambitious 3D project but was uneven, while developer Theo Browne criticised its handling of long and simple tasks. OpenAI says Astra performs better than Claude Fable 5.2 on its Agent’s Last Exam at a lower cost; independent figures cited in the article put Astra’s Intelligence Index score at 62 and cost at $4 per task, compared with Fable’s 65 and $26.54. GPT-6.2 has not yet been assessed by the Remote Labor Index, where the cited automation rates are 20.83% for GPT-6 Astra and 41.25% for Fable 5.2. CEO Sam Altman acknowledged usability concerns and said improving how the models are to work with would be a priority.

---
Summarised by TalkFuse (https://talkfuse.com/). Quote the original publisher for the reporting and link this page for the summary and discussion.
