GPT-5.5 has only been released for 3 weeks, and GPT-5.6 has already been run by external developers in the Codex environment. Is OpenAI's current iteration pace on a monthly basis? According to the developer's actual test data in the OpenCode environment, the context window of the new model reaches 1.5M tokens, which is about 43% higher than the official setting of 1.05M for GPT-5.5 API. The call trace of this model first appeared in the Codex routing log on April 28th. After entering this week, the "model not supported" error that previously occurred with gpt-5.6 has disappeared, replaced by normal system response. Tests have shown that the model can stably process requests for over 900000 to 1.05M tokens, self report the running environment as openai/gpt-5.6 during conversations, support xhigh inference level, and provide fast mode with high response speed.