OpenAI GPT-5.6 Multi-Agent V2: 16X Speed Boost, 741 Convos/Sec

OpenAI has officially launched GPT-5.6 Multi-Agent V2, significantly boosting its speed and capabilities. This update allows for 741 rounds of extreme heavy-load conversations per second. An internal Slack message revealed an "epic" full-scale performance overhaul for ChatGPT. Application loading speed has soared by 94%, and memory usage is down by 41.2%. The number of network requests plummeted by 98.2%, and conversation history loading time dropped by 99.6%. The new Multi-Agent v2 version of Codex enables the main agent to automatically delegate subtasks to different models. Each sub-agent can have its own reasoning intensity settings. OpenAI President Greg Brockman stated they are moving away from manual model selection. This upgrade aims to eliminate waiting and selection for users, delivering a faster, smoother interactive experience. For a "monster-level" conversation of 741 rounds and 231MB, the average opening time has dropped from over 27 seconds to under 2 seconds. The number of network requests for such a conversation has been cut from 894 to just 16. This means a much more efficient and responsive experience for users engaging in complex, long-form interactions.