- By Prateek Levi
- Mon, 17 Aug 2026 12:13 PM (IST)
- Source:JND
OpenAI is testing a new Ultrafast mode for GPT-5.6 Sol, and the headline feature is pretty straightforward: speed. The new mode is currently being offered as a limited preview to a small group of customers.
OpenAI says GPT-5.6 Sol can run up to 14 times faster with Ultrafast than it does with standard processing. The mode uses Cerebras hardware and can reportedly generate up to 750 output tokens every second.
ALSO READ: BGMI Lite Coming To India By End Of 2026: Here’s What To Expect
For businesses, that extra speed could make a noticeable difference. Instead of waiting for an AI system to work through a task, companies could get responses much more quickly, which could be useful when AI is being used as part of a live workflow.
OpenAI is looking at several areas where this could come in handy, including real-time incident response, customer support, market analysis, transaction assessment and live research. The company is also testing the technology with businesses involved in coding, commerce and financial research.
The company says these early customers are helping it figure out where a big increase in AI speed is actually useful, rather than simply making responses arrive faster. What OpenAI learns from these tests will also help determine how it expands Ultrafast in the future.
OpenAI is testing the mode internally as well. Its teams are using it for incident response, where a faster model can potentially cut down the time between spotting a problem, investigating it and deciding what to do next.
Researchers are also trying Ultrafast for tasks that involve searching knowledge sources, querying data and pulling information from connected tools before putting everything together into a useful summary.
For now, there’s no broad rollout. OpenAI says Ultrafast is available only to selected customers, with access expected to expand as more capacity becomes available.