How fast is Ultrafast mode compared to normal?

OpenAI says it can run GPT-6 Astra up to 8 times faster than normal, at up to 300 tokens per second. An earlier Ultrafast preview ran GPT-5.6 Sol up to 14 times faster at up to 750 tokens per second.

Which subscription plan do I need for Ultrafast?

It requires the new top-tier Pro 500 plan or the pay-as-you-go API. Pro 500 has 25 times the usage cap of the standard Plus plan.

What does Ultrafast cost on the API?

Running Astra in Ultrafast mode on the API costs 6 times more than running it at normal speed.

Large Language ModelsAI Business & IndustryITmedia AI＋12h agoSep 30, 2026, 04:00 JST

OpenAI's Ultrafast runs GPT-6 Astra up to 8x faster

Send on LINE

3 Key Points

What happened
OpenAI announced Ultrafast at its OpenAI DevDay 2026 conference. It runs GPT-6 Astra up to 8 times faster than normal, and GPT-6.1 Sol will get the mode soon.
Why it matters
The speed boost could mean the same model that once took longer to answer can now respond faster, potentially making Astra more practical for tasks that need quick output.
What to watch
Access hinges on OpenAI's new top-tier subscription plan or metered API, not the standard paid option. Watch whether the pricing gap affects who can actually use it.

WHO IT HITSDevelopers and teams building on OpenAI's API will likely weigh the 6x price premium for Ultrafast Astra against the speed gain, while ChatGPT Work and Codex subscribers on the top-tier plan get the mode as a bundled feature.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

OpenAI introduced Ultrafast for GPT-6 Astra at its OpenAI DevDay 2026 conference, saying the mode runs the model up to 8 times faster than normal. The same conference also produced GPT-6.1 Sol, which OpenAI says will get Ultrafast access soon. In subscription settings, the mode can be turned on in ChatGPT Work and Codex and is described as reaching up to 300 tokens per second.

Ultrafast is not entirely new. It first appeared in August as an API offering for GPT-5.6 Sol, then billed as up to 14 times faster at up to 750 tokens per second, and it uses Cerebras inference processors on the backend. Alongside the mode, OpenAI restructured its paid tiers: the new Pro 500 plan has 25 times the usage cap of the standard Plus plan, the former Pro 20x became Pro 200 with its cap halved to 10 times, and the former Pro 5x became Pro 100 with its 5-times cap unchanged. API users pay 6 times the normal rate to run Astra at Ultrafast speed.

What the mode ultimately changes for users appears to hinge on how much the speed gain is worth relative to the higher price and the tier reshuffle, particularly for API-dependent work versus subscribers on the top plan.

FAQ

How fast is Ultrafast mode compared to normal?: OpenAI says it can run GPT-6 Astra up to 8 times faster than normal, at up to 300 tokens per second. An earlier Ultrafast preview ran GPT-5.6 Sol up to 14 times faster at up to 750 tokens per second.
Which subscription plan do I need for Ultrafast?: It requires the new top-tier Pro 500 plan or the pay-as-you-go API. Pro 500 has 25 times the usage cap of the standard Plus plan.
What does Ultrafast cost on the API?: Running Astra in Ultrafast mode on the API costs 6 times more than running it at normal speed.

ITmedia AI＋Read Original Article

Send on LINE

Also reported by GIGAZINE AI

AI news that matters for your work, in one minute a day

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.