ChatGPT frontend speed up 94%, Multi-Agent v2 automates model selection

新智元 · wechat · 2026-08-16

OpenAI has achieved a massive performance boost for ChatGPT, reducing load times for long conversations by 94%, cutting network requests by 98.2%, and lowering memory usage by 41.2%. Simultaneously, the Multi-Agent v2 system is fully live, allowing the primary Agent to automatically delegate subtasks to different models (e.g., Sol, Terra, Luna) with individualized reasoning intensity settings. This shift automates model selection, reduces inference costs, and accelerates ChatGPT's evolution into a fully automated workflow platform.

Original post →

More from coding & agent

coding & agent channel →