OpenAI's newly launched GPT-6 Astra is facing a wave of user complaints just a week after its debut, with many claiming the model has become noticeably less capable than it was at launch — echoing a pattern that played out with the company's previous release earlier this year.
Users Report a Decline in Performance
Across social media and community forums, subscribers who initially praised GPT-6 Astra say its responses have grown shorter, less detailed, and less accurate over the past several days. Some describe the model as slower to reason through complex prompts, while others report it failing at tasks it handled with ease during its first days of availability.
The complaints follow a familiar script within the AI community, where the term "nerfed" — borrowed from gaming — has become shorthand for the perception that a model has been quietly weakened after its splashy launch.
Users say the model that wowed them at launch now feels like a shadow of its former self.
For many paying customers, the frustration is compounded by a lack of transparency. Without clear documentation of what, if anything, has changed under the hood, users are left comparing their own before-and-after experiences to determine whether the model truly regressed.
A Pattern OpenAI Has Seen Before
This is not the first time OpenAI has weathered such criticism. The company's previous flagship model went through an almost identical cycle in July, when users voiced similar concerns that performance had dropped in the weeks following release.
Whether these perceived declines reflect actual changes to the model or shifting user expectations remains a point of debate. AI companies often adjust systems after launch to manage computing costs, reduce latency, or tighten safety guardrails — any of which can alter how a model behaves.
Several factors could explain the recurring pattern of post-launch disappointment:
- Behind-the-scenes tuning to reduce server load and operating expenses
- Added safety filters that make responses more cautious or restrictive
- The novelty effect wearing off as users push the model harder over time
For now, OpenAI has not publicly confirmed any specific modifications to GPT-6 Astra. Until it does, users are likely to keep scrutinizing every response, wondering whether the model they signed up for is the same one they're getting today.