AI product
Ultrafast is described in the videos as OpenAI's high-speed inference mode for Codex, intended for real-time, in-the-loop coding. The videos say it runs models such as Astra at more than 300 tokens per second, while charging $300 per million output tokens—about six times standard pricing—and discuss a $500-per-month plan.
1 use taken from transcripts — each links to the moment in the video.
OpenAI's ultra-fast inference mode for Codex, running models like Astra at 300+ tokens per second for real-time, in-the-loop coding. The video reviews its extremely high cost ($300/million output tokens, 6x standard pricing) and a $500/month plan it says is not worth it.
1 in the library.