OpenAI Previews Ultrafast Mode For GPT-5.6 Sol, Running Up To 14X Faster

OpenAI has previewed Ultrafast, a new service tier that can run GPT-5.6 Sol up to 14 times faster than Standard processing and generate up to 750 output tokens per second, targeting enterprise workflows where every second matters.

The new mode is launching first through the OpenAI API and is currently available only to a select group of customers as part of a limited preview. It is powered by Cerebras, extending the two companies' partnership around low-latency artificial intelligence (AI) inference.

TL;DR

 

• OpenAI says Ultrafast runs GPT-5.6 Sol up to 14X faster than Standard processing.
• The service can generate up to 750 output tokens per second.
• Ultrafast launches first through the OpenAI API and is powered by Cerebras.
• OpenAI is testing it across coding, finance, support, commerce, research, and incident response.
• Access remains limited while OpenAI expands capacity.

OpenAI announced Ultrafast on August 13, describing it as a new speed class for its most capable GPT-5.6 model.

“Until now, getting real-time speed typically meant choosing a smaller or more specialized model,” OpenAI said in its official announcement, adding that Ultrafast represents a move toward delivering “more useful work per second.”

TechCrunch also reported that the mode is designed to substantially accelerate GPT-5.6 Sol, while noting that the company's headline performance figures reach up to 14X Standard processing speeds and 750 output tokens per second.

Source

It is worth stressing the “up to” part. OpenAI is presenting 14X as a maximum performance figure rather than claiming that every Ultrafast request will always finish fourteen times faster.

What Can GPT-5.6 Sol Ultrafast Be Used For?

The idea is to bring OpenAI's higher-end intelligence into workflows where latency can limit how useful an AI model is.

OpenAI listed incident response and reliability among the potential applications. During a system failure, Ultrafast could analyse application logs, recent code changes, traces, and engineering reports quickly enough to help teams identify likely causes and prepare fixes while an incident is still unfolding.

The company also highlighted financial research and security, customer support and voice applications, commerce, and live research and experimentation. For example, an AI-powered customer service system could handle multi-step questions without creating lengthy pauses in a live conversation, while researchers could potentially turn some workflows that previously ran overnight into interactive sessions.

9to5Mac similarly noted that OpenAI sees the technology supporting latency-sensitive workloads spanning voice, commerce, developer agents, financial research, customer support, and security response.

OpenAI said its own developers are already experimenting with the mode. Its engineering teams have used Ultrafast to read logs, analyse traces, synthesize conversations, identify follow-up checks, and help prepare or validate fixes during incident response.

The company stressed that engineers remain responsible for judgment and deployment decisions.

Cerebras Powers OpenAI's New Speed Tier

The performance boost comes through OpenAI's partnership with Cerebras, which is providing the infrastructure behind GPT-5.6 Sol's Ultrafast mode.

According to OpenAI, Cerebras is supporting the model at rates of up to 750 output tokens per second. The company says this could enable businesses to create more responsive applications and introduce advanced AI into workloads where inference delays previously made using a frontier model less practical.

Daily.dev also highlighted Cerebras as the underlying technology provider and reported the same headline performance of up to 14X faster inference and up to 750 output tokens per second. It additionally pointed to the Cerebras hardware architecture as a central part of the speed-focused deployment.

The partnership reflects a broader challenge for AI developers. Faster responses have often involved choosing smaller or more specialized models, whereas OpenAI is positioning Ultrafast as a way to retain GPT-5.6 Sol's capabilities while significantly reducing the time users spend waiting for output.

What Are Early Ultrafast Customers Saying?

OpenAI said an initial group of businesses is testing GPT-5.6 Sol Ultrafast across coding, financial research, commerce, support, and other interactive applications.

“The increase in speed brought by Cerebras is impressive,” said John Crepezzi of AI Assistants at Jane Street. He said the improvement makes new ways of working alongside the models practical for developers.

Courtland Lykins, Product Lead for Voice AI at Podium, said Ultrafast had been valuable for the company's voice stack, particularly for more complex work where response speed affects the call experience.

When Will OpenAI Ultrafast Be Widely Available?

Not yet.

GPT-5.6 Sol Ultrafast is currently available only as a limited preview for a select group of customers. OpenAI said it will broaden availability as capacity expands and is inviting interested businesses to sign up for access updates.

 

TechCrunch and 9to5Mac likewise reported that access remains restricted during the preview rather than representing a general rollout to all API or ChatGPT users.

For now, Ultrafast represents OpenAI's attempt to make response speed a feature of frontier-level AI rather than a tradeoff that requires moving to a smaller model. Whether its maximum performance translates consistently across real production workloads should become clearer as OpenAI expands the preview.