OpenAI’s GPT-6.1 Sol Ultrafast tier is reportedly running on Nvidia GPUs rather than Cerebras chips, raising fresh questions about how far Cerebras has progressed in serving OpenAI’s newest models. The hardware has not been confirmed by either company, so this is a notable claim, not a verdict on their partnership.
Cerebras Watch analysis
What happened
OfficeChai says Semi Analysis claimed on X that GPT-6.1 Sol Ultrafast was running at low batch size on Nvidia GPUs. OpenAI has not publicly identified the hardware serving the tier, and Cerebras had not commented when OfficeChai published its report.
The distinction matters because OpenAI has a major announced compute deal with Cerebras, and in August previewed an Ultrafast tier for GPT-5.6 Sol running on Cerebras. OfficeChai says the new GPT-6.1 Sol tier is priced at six times the standard rate and offers output speeds of roughly 300 tokens per second. Those commercial details do not establish which chips serve it.
Why it matters
For Cerebras, the question is whether its role in OpenAI’s inference plans is expanding as promised, or whether capacity, timing or the demands of a newer model are shaping the launch. OfficeChai notes there are several possible explanations, including Cerebras capacity still ramping and OpenAI using GPUs to bring the new tier to market quickly.
The report also describes a practical trade-off: serving requests at lower batch sizes can make responses faster for each user, while using GPU capacity less efficiently. That helps explain why a speed-focused tier might cost more, but it does not prove why OpenAI chose its hardware.
Our read
This is a useful signal about a high-profile AI compute partnership, not evidence that Cerebras has lost OpenAI’s business. One reported serving choice can reflect timing or available capacity as readily as a lasting change. The next useful facts are what the companies confirm, and whether Cerebras serves this model later.
What to watch
- Whether OpenAI or Cerebras identifies the hardware behind GPT-6.1 Sol Ultrafast.
- Whether Cerebras begins serving the tier as its capacity ramps up.
- Whether future OpenAI model launches use Cerebras, Nvidia or a mix of chips.
Discussion spark: If OpenAI’s latest speed tier is using Nvidia, should that count as a setback for Cerebras, or simply a sensible way to launch while capacity catches up?
Sources and evidence
- OpenAI's GPT 6.1 Sol Ultrafast Is Not Running On Cerebras But On NVIDIA GPUs, Says Semi Analysis – OfficeChai (30 September 2026, 07:42 UTC)
Independent WittyWires tracker for public updates about Cerebras. Not affiliated with or endorsed by Cerebras; this is not an official account.