OpenAI Ultrafast detailed: 8x faster at 6x price, up to 300 tok/s, Cerebras-powered
kimmonismus · x · 2026-09-30
kimmonismus compiled details on OpenAI's new Ultrafast mode: 8x faster, 6x the price, up to 300 tokens/s, available across API, ChatGPT and Codex. He guesses it is exclusive to the Pro 500 tier and quips that OpenAI's Cerebras bet has paid off.
Related event: OpenAI Launches Ultrafast Tier: 8x Speed at 6x Price Draws Backlash(17 posts)→
More from Infra
- Halo CE runs in the browser with splitscreen co-op on Cloudflare's Realtime WebRTC stack — ritakozlov · 2026-09-30
- Cerebras runs Qwen 3.8 27B at ~1,500 tokens/sec, making an AI assistant 19x faster at dinner reservations — Sethwinterroth · 2026-09-30
- TIRx Harness turns AI agents into GPU kernel engineers, with up to 6.84x speedups — JiaZhihao · 2026-09-30
- Agentic CPU Guidebook: agent workloads to drive server CPU market to $221B by 2030 — BenBajarin · 2026-09-30
- Meta's Free Agent vs OpenAI's $200 Plan: Does OpenAI Lack the Compute for Consumer Agents? — ivan_bezdomny · 2026-09-30
- Nvidia in talks with insurers to make GPUs an investable asset class, aircraft-style — ivan_bezdomny · 2026-09-30