2 weeks ago
OpenAI's Ultrafast Mode Makes GPT-5.6 Sol 14 Times Faster
OpenAI has a very smart computer program called GPT-5.6 Sol that can answer questions and write text.
This program is really good, but sometimes it is slow because it has to think hard.
OpenAI made a new special setting for it called 'Ultrafast' that lets it answer up to 14 times faster.
When it works fast, it can create about 750 tiny word-pieces, called tokens, in just one second.
That is a lot of writing in a very short time.
The speedy setting works with help from another company called Cerebras, which builds special computers for fast thinking.
Right now, only certain companies can try the fast setting through a special connection called an API.
Companies like Jane Street and Podium are testing it to see how much it helps them.
OpenAI is also making its AI faster because other companies have fast AI too, and it wants its AI to stay one of the best.
OpenAI announced 'Ultrafast,' a new service tier that runs GPT-5.6 Sol up to 14 times faster than standard processing.
The mode can generate up to 750 output tokens per second and is powered by technology from AI hardware company Cerebras.
'Ultrafast' is available as an early preview via the API, aimed at power users who want ChatGPT speed without switching to smaller models.
The feature is being tested with companies including Jane Street, Podium, Basis, and Rogo across programming, commerce, financial research, support, and other interactive applications.
OpenAI has not disclosed per-token costs, and the launch comes as US labs face competition from Chinese open-weight models such as Moonshot's Kimi K3 and rivals like Anthropic's Claude fast mode.
- Who
- OpenAI, a US AI lab, together with its own developers and initial test customers including Jane Street, Podium, Basis, and Rogo.
- What
- Launched 'Ultrafast,' a new service tier for GPT-5.6 Sol that runs the model up to 14 times faster and generates up to 750 output tokens per second.
- Where
- Not explicitly stated; OpenAI is described as a US AI lab and the feature is accessed through its API.
- When
- Announced in a recent OpenAI blog post; no specific date was given in the article.
- Why
- To let power users get cutting-edge performance at much higher speeds without resorting to smaller models, and to compete with faster options from rivals such as Anthropic and with Chinese open-weight models like Moonshot's Kimi K3.
Key facts
- Company
- OpenAI
- Feature
- 'Ultrafast' service tier for GPT-5.6 Sol
- Speed increase
- Up to 14 times faster than standard processing
- Output speed
- Up to 750 output tokens per second
- Hardware partner
- Cerebras
- Availability
- Early preview via API; companies can join a waitlist
- Test customers
- Jane Street, Podium, Basis, Rogo
- Pricing
- Per-token costs not disclosed
Quotes
OpenAI
OpenAI’s official statement
“"Ultrafast" can generate up to 750 output tokens per second.”
thehansindia.com
Sources
Related news
OpenAI Cuts GPT-5.6 Sol Prices After AI Test Pause
ChatGPT Rolls Out GPT-5.6 Luna, Unlimited Chats and Think Button
OpenAI Reports Jalapeño Chip Gains in Speed and Efficiency
OpenAI Cuts GPT-5.6 Sol Developer Prices Amid Competition
OpenAI Cuts Frontier GPT-5.6 Sol Developer Prices Over 20%
SpaceXAI unveils Grok 4.6, claims frontier-level performance at lower cost
China Challenges US Leadership in Escalating Global AI Contest
Google's Gemini 3.7 Flash Gains Speed Before Price Doubles
Trump-backed crypto firm backs AI venture offering restricted Chinese models
Alibaba plans revenue share for users of Qwen open-source model
