Frontier models advance.

OpenAI released GPT-5.1 this week, improving multi-step reasoning, planning, and code generation across ChatGPT and its developer tools.

Recent research on why models hallucinate is starting to show itself in the models' outputs, with answers that not only score higher but also deliver improved performance. Model training is complex, and users can see how well these systems are improving at handling context, accuracy, and autonomous tasks.

GPT-5.1 follows not long after the release of Kimi 2 Thinking, where both models shift towards efficiency and architecture rather than just increasing raw parameters.

Read the paper on hallucinations here: https://openai.com/index/why-language-models-hallucinate/