OpenAI Launches Ultrafast: GPT-5.6 Sol Hits 750 Tokens/sec
量子位 · wechat · 2026-08-14
OpenAI announced the Ultrafast preview tier, accelerating GPT-5.6 Sol by up to 14x to 750 tokens/sec without compromising intelligence. This leverages Cerebras wafer-scale engines to bypass traditional memory bandwidth bottlenecks.
Additionally, the ChatGPT desktop app introduced Computer History. It tracks user interactions like clicks, keystrokes, and app switches, allowing the AI to answer context-aware questions like "where did I leave off?" without taking screenshots or recording audio. Raw data is kept locally for 48 hours and not used for training.
Related event: OpenAI and Cerebras Preview Ultrafast Mode for GPT-5.6 Sol at 750 Tokens/s(20 posts)→
More from Infra
- Qwen3.8-27B on 24GB VRAM: 131k Context with MTP Enabled — sisyphus-cycle · 2026-08-17
- Running dstack Confidential VMs for private code execution on cloud — bgmshana · 2026-08-17
- AI for hardware engineering: Can models understand and improve complex structures? — rms80 · 2026-08-17
- US per capita power consumption peaked at dot-com,暗示 scaling limits — jwt0625 · 2026-08-17
- User runs Krea2 and MiniMax Music locally — -becausereasons- · 2026-08-17
- MiniMax H3 Video Generation Tested on RTX 3060 12GB — solomars3 · 2026-08-17