OpenAI has acknowledged problems with the ChatGPT Work and GPT-5.6 Sol launch, The Decoder reported. The issues include excessive compute use, a confusing transition to desktop workflows, and unclear distinctions between product modes.
The admission is notable because enterprise AI products need predictable costs and understandable user flows. Even strong models can struggle if customers cannot tell which mode to use or how much a workflow will consume.
The episode shows that AI product quality now depends on packaging, pricing, and workflow design as much as model capability.