Reports indicate that OpenAI has recently rolled out a new-generation coding model, GPT-5.1-Codex-Max. This model is purpose-built to tackle long-duration, highly complex engineering tasks. It boasts capabilities such as cross-session planning, reasoning, and code revision. The core innovation of this model is the introduction of a 'compression' mechanism. When the model nears the context window limit, this mechanism enables it to intelligently retain crucial information while discarding irrelevant details. As a result, it can maintain contextual coherence in million-token tasks and operate steadily for more than 24 consecutive hours. In numerous programming benchmark tests, GPT-5.1-Codex-Max has showcased outstanding performance. For instance, in the SWE-Bench Verified test, it achieved an accuracy rate of 77.9%, surpassing Google's Gemini 3 Pro, which scored 76.2%. At present, OpenAI has integrated this model into multiple development environments, making it available for developers to utilize.