User feedback on GPT-6.2 has split sharply since launch. Some users complain it takes hours of repeated back-and-forth to get a satisfactory result, and that long tasks get cut off by the conversation length limit. AI commentator Nate B Jones offers a different reading: a model obsessed with detail and highly token-efficient, whose strengths simply do not line up with what most people expect.

The defense: a detail-obsessed model built for agent communication

Against the wave of negative reviews, Nate B Jones offers a different interpretation. He calls GPT-6.2 a "very smart but can't-see-the-forest" model, fixated on details and repeated verification, with extremely high token efficiency — a trait that matters a lot to OpenAI as it runs back into compute constraints.

Nate B Jones traces the bad reviews to two sources: weak writing ability, and strong performance on "tree-shaped" problems but poor performance on "forest-level" engineering problems. That same trait, he argues, makes the model unusually well suited to agent-to-agent communication — high information density, heavy emphasis on verification.

It is a very smart but "can't-see-the-forest" model, fixated on details and repeated verification, with extremely high token efficiency.

Hands-on tests: high-quality output, high friction

Blogger imjustnewatai said he wanted to share more hands-on results with GPT-6.2, but each good result took hours of back-and-forth with the model, limiting how often he could post. That suggests the model can produce high-quality output, but controllability and stability still come with obvious friction.

Long tasks: the conversation length limit cuts key workflows

Another user posted a ChatGPT screenshot: he asked the model to "compose a symphony and shoot a stunning video around the score." The model had barely started planning when it showed "Stopped working." The user clicked Continue to ask why, and got a prompt saying the conversation had hit its maximum length limit and a new conversation was needed to continue. The poster used it to vent that "this is why I hate OpenAI" — a long task cut off at a key moment by the conversation length limit, with no way to carry the context forward. The experience, he said, was deeply frustrating.

What to watch

Taken together, the GPT-6.2 controversy centers on two dimensions: model style and controllability. Its detail orientation and token efficiency are advantages in settings such as agent communication, but writing, long-horizon tasks and interaction stability remain clear weak spots. Worth watching next: whether OpenAI adjusts the conversation length limit and the model's controllability, and whether user feedback pushes the company to define more clearly where the model is meant to be used.