Cerebras Powers GPT-5.6 Sol Ultrafast in OpenAI API at Up to 750 Tokens per Second
Cerebras and OpenAI are giving an early look at Ultrafast Mode, a new OpenAI API tier that runs GPT-5.6 Sol on Cerebras hardware at up to 750 output tokens per second. Coverage cites up to 14x faster inference than standard options, while Cerebras says there is no quality compromise and points to benchmarks versus Fable 5 and Opus 4.8. The Decoder frames Standard, Fast, and Ultrafast as separate tiers under the companies' reported $10 billion partnership. OpenAI is offering Ultrafast first to a select API customer set, with broader access tied to growing capacity. The preview matters because it productizes raw throughput for enterprise and real-time agent workloads, though availability remains invite-limited and external validation of speed and quality claims is not yet shown in the packet.
Cerebras Powers GPT-5.6 Sol Ultrafast in OpenAI API at Up to 750 Tokens per Second
Today, Cerebras and OpenAI are sharing an early look at Ultrafast Mode, a new service tier launching first in the OpenAI API and powered by Cerebras. Cerebras powers GPT-5.6 Sol on Ultrafast mode, delivering up to 750 output tokens per second and without any quality compromise. Ultrafast is available initially to a select group of customers, with access expanding over time.
Key takeaway
Frontier-model competition is shifting toward monetized inference speed tiers backed by specialized silicon, not only model capability upgrades.
What happened
Cerebras and OpenAI shared an early look at Ultrafast Mode, a new service tier launching first in the OpenAI API and powered by Cerebras hardware for GPT-5.6 Sol.
Reporting attributes up to 750 output tokens per second and up to 14x faster runs versus standard options, with initial access limited to a select customer group before broader rollout.
Evidence
Cerebras and OpenAI previewed Ultrafast Mode as a new OpenAI API tier powered by Cerebras.
Cerebras Blog · attributed
Today, Cerebras and OpenAI are sharing an early look at Ultrafast Mode, a new service tier launching first in the OpenAI API and powered by Cerebras.
Ultrafast mode targets up to 750 output tokens per second without a stated quality compromise.
Cerebras Blog · attributed
Cerebras powers GPT-5.6 Sol on Ultrafast mode, delivering up to 750 output tokens per second and without any quality compromise.
The Decoder reports Ultrafast is powered by Cerebras under a $10 billion partnership and sits alongside Standard and Fast tiers.
The Decoder · attributed
OpenAI is launching "Ultrafast," a new inference mode that delivers GPT-5.6 Sol at up to 750 output tokens per second, powered by Cerebras hardware from their $10 billion partnership.
TechCrunch reports OpenAI launched Ultrafast in preview for GPT-5.6 Sol at up to 14x speed and 750 tokens per second.
TechCrunch AI · attributed
OpenAI has launched 'Ultrafast,' a new processing mode for GPT-5.6 Sol that delivers up to 14x speed and 750 tokens per second.
OpenAI said Ultrafast is rolling out first in the API to a select customer group.
OpenAI (X) · attributed
Previewing Ultrafast mode: GPT-5.6 Sol at up to 14x the speed. Launching first in the OpenAI API to a select group of customers with expanded access to more businesses as capacity grows.
Why it matters
Teams building latency-sensitive voice or interactive agents must compare Standard, Fast, and Ultrafast pricing against measurable end-user responsiveness gains.
Limits and uncertainties
Ultrafast is available initially to a select group of customers, with access expanding over time.
Cerebras claims delivery without any quality compromise, but the packet provides no independent third-party validation of that claim.
Practical implications
Evaluate whether Ultrafast latency justifies a new API tier cost for real-time product surfaces before committing integrations.
Plan architecture around tiered inference options rather than assuming a single GPT-5.6 Sol latency profile.
What to watch
OpenAI expansion of Ultrafast API access beyond the initial select customer cohort as capacity grows.
Published pricing and benchmark details comparing Standard, Fast, and Ultrafast tiers under production workloads.
