Every story we've covered involving model-performance.
Users on X and AI developers are reporting a notable decline in GPT-6 Astra's output quality roughly a week after its release, with many calling the model "dumber" or showing worse code and reasoning compared with launch-day results. The backlash echoes a similar user outcry last July over OpenAI's previous flagship, GPT-5.6 Sol, which also saw complaints about a sudden drop in its top reasoning mode.
No stories here yet.