OpenAI Launches Ultrafast Speed Tier: 8x Faster, 6x Pricier, Alongside Pro 500 Plan
On 09-30, OpenAI launched a high-speed inference tier called Ultrafast, for the first time offering it across its entire product line: it's available via API, ChatGPT, and Codex.
Confirmed
- Speed: token generation in Codex is up to 8x faster (300 tokens/second), and up to 6x faster via the API; multiple posters (@cneuralnetwork, @YashasGunderia, @kimmonismus) consistently confirmed roughly an 8x speedup, reaching about 300 tokens/s.
- Pricing: the Ultrafast tier costs roughly 6x the standard mode.
- Coverage: @YashasGunderia said the tier is rolling out across six entry points spanning API, ChatGPT, and Codex; @kimmonismus noted that Ultrafast versions of GPT-6 Astra and GPT-6.1 Sol are already live.
Unconfirmed
- @kimmonismus speculated that Ultrafast's speedup is powered by Cerebras hardware; this is the poster's personal guess and was not directly confirmed in official posts.
Why it matters
- This is the first time OpenAI has offered an ultra-low-latency tier as a standard option across its entire product line—ideal for latency-sensitive use cases, though users must weigh the speed-versus-cost tradeoff.
2026-09-30 ~ 2026-09-30 · 16 related posts
Primary sources
- OpenAI's New Ultrafast Tier: Up to 8x Faster Codex at 300 Tokens/sec for Pro 500 Plans — testingcatalog · 2026-09-30
- OpenAI's Ultrafast mode: 8x faster, 6x the price — rudrank · 2026-09-30
- GPT-6.1 Sol Ultrafast Goes Live: Up to 300 token/s, 8x Faster at 6x the Price — kimmonismus · 2026-09-30
- OpenAI launches Ultrafast across API, ChatGPT, and Codex — YashasGunderia · 2026-09-30
- All models get ultrafast: ~300 tps, 8x faster but 6x more expensive — cneuralnetwork · 2026-09-30
- OpenAI launches Cerebras-powered ultrafast tier: 8x speed for 6x price — downingARK · 2026-09-30
- OpenAI Ultrafast detailed: 8x faster at 6x price, up to 300 tok/s, Cerebras-powered — kimmonismus · 2026-09-30
- OpenAI Ultrafast mode announced: 8x speed, 6x price, crowd boos — ethotopia · 2026-09-30
- Hands-on demo shows OpenAI's Ultrafast is impressively fast at work — testingcatalog · 2026-09-30
- [source] OpenAI launches Ultrafast: up to 8x faster tokens at 300 tok/s in Codex — OpenAI · 2026-09-30
- [source] OpenAI's Ultrafast goes live with Pro 500 plan offering 25x Plus usage limits — OpenAI · 2026-09-30
- [source] OpenAI launches Codex Ultrafast: up to 8x faster than Astra Standard — OpenAIDevs · 2026-09-30
4 near-duplicate retellings: downingARK · xiaohu · OpenAIDevs · OpenAI