OpenAI Unveils Ultrafast Mode for GPT-5.6 Sol With 14x Faster AI Responses

The accuracy of AI models is steadily increasing. Among the most cutting-edge AI models available to users now is OpenAI's GPT-5.6 Sol. Frontier models take time to process requests, which is a concern. But now OpenAI claims to have the answer; their new Ultrafast mode for Sol makes the model fourteen times quicker.

According to a blog post by OpenAI, a new Sol service tier called Ultrafast can execute GPT5.6 Sol up to fourteen times quicker than Standard processing. "More useful work per second" is what the business claims Ultrafast will make possible.

Offerings of Ultrafast Mode

Ultrafast can supposedly produce 750 output tokens/second, according to OpenAI. To those who are not aware, tokens serve as a fundamental unit of measurement for AI models. The more work users perform, the more tokens they consume. Anyone who has wished for faster ChatGPT without resorting to smaller AI models can find what they're looking for with Ultrafast. According to OpenAI, the mode is driven by Cerebras, an artificial intelligence hardware startup that specialises in ultra-low latency AI inference. The business claims that Ultrafast is ideal for jobs that require lightning-fast processing times.

In the midst of an outage, it can also be useful for analysing application logs, code modifications, and engineer reports. On top of that, it can handle consumer problems instantly and evaluate transactions and suspicious activity even when factors change. When an event occurs, the developers at OpenAI use Ultrafast to scan logs, analyse traces, synthesise discussions, determine next checks, and assist in fixing the problem.

Also, OpenAI's research loops, which used to run overnight, are now much more tightly wound thanks to Ultrafast. This change is happening at a moment when OpenAI and other US AI laboratories are up against formidable competition from locally executable open-weight models developed in China, such as Moonshot's Kimi K3. Ultrafast, according to OpenAI, is intended to provide frontier-level performance at substantially greater speed, particularly since Anthropic and other OpenAI rivals have introduced speedier choices like Claude's fast mode.

Ultrafast Going Through Rigorous Testing

As of right now, the capability is only available in preview via API, according to the business. Throughout the trial period, OpenAI is collaborating with a select number of customers to identify the areas where the increased speed has the most impact on real-world goods. Companies such as Jane Street, Podium, Basis, and Rogo are testing Ultrafast in their interactive apps, which span coding, commerce, financial research, and support.

Companies can get on the Ultrafast waitlist by providing information like workload, latency needs, and anticipated usage. OpenAI will then assess the early deployment and get ready to open it up to more people. Ultrafast's token costs are currently unknown, according to OpenAI.