OpenAI is rolling out Ultrafast mode for GPT-6.1 Sol in the API, Codex and ChatGPT Work, according to an announcement in the OpenAI Developer Community on 8 October 2026. OpenAI describes it as near-Astra intelligence at up to 8x the speed of Sol Standard.
API pricing for GPT-6.1 Sol in Ultrafast mode is $12 per million input tokens and $60 per million output tokens, which OpenAI says is 1.2x the cost of Astra. OpenAI’s pricing page lists Sol’s Standard rate at $2 and $10, so Ultrafast costs six times as much per token on our arithmetic.
Update, 6 Nov, 05:25 UTC: OpenAI is ending Ultrafast for GPT-5.6 Sol, the tier Cerebras powered. OpenAI told API customers it is “deprecating ultrafast for GPT-5.6-Sol in the API on Dec 7, 2026”, according to an OpenAI email shown in a post by SemiAnalysis at 04:59 UTC on 6 November. After that date, requests for gpt-5.6-sol on the ultrafast tier “will fall back to the standard service tier”. The model stays; its Ultrafast tier ends. The email points customers to Ultrafast on GPT-6.1 Sol and GPT-6 Astra as replacements.
When OpenAI previewed Ultrafast for GPT-5.6 Sol, it said the tier was “Powered by Cerebras” and generates “up to 750 output tokens per second”. For the replacement, SemiAnalysis said on X on 29 September that GPT-6.1 Sol Ultrafast was not running on Cerebras but at a low batch size on NVIDIA GPUs, as reported by OfficeChai; as of that report, OpenAI had not said which hardware serves GPT-6.1 Sol Ultrafast.
Cerebras’s 10-Q for the quarter to 30 June describes a December 2025 agreement for 750 MW of inference capacity that OpenAI “is contractually committed to purchase”, which the filing frames as capacity, not a particular model. Cerebras recognised $56.8 million of revenue under it in that quarter, about 31.5% of its $180.1 million total (our arithmetic), and said “a significant amount” of its $25.4 billion of remaining performance obligations came from the agreement. Our megawatt-deal piece sets out its structure. Cerebras (CBRS) closed at $81.71 on 5 November, down 11.0% on the day per Nasdaq’s daily history, before the notice was posted. We found no comment from Cerebras or OpenAI by 05:35 UTC.
Business Pill · INSTANT OR CHEAP
A short explainer of the trade-off between latency and throughput: why an AI system cannot be both instant and cheap. It teaches the general idea only and says nothing about any company in this story.
The key insight: As we read it, OpenAI now prices speed separately from capability on the same model. GPT-6.1 Sol costs $2 and $10 per million tokens at Standard and $12 and $60 at Ultrafast, a sixfold step on our arithmetic for the same model run faster.
The Price of Speed
On OpenAI’s API pricing page, GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens in Standard mode, $4 and $20 in Fast mode, and $12 and $60 in Ultrafast mode, for prompts of up to 272K tokens. Above that, Ultrafast is $24 and $90.
OpenAI positions the mode for work where speed and intelligence both matter, naming debugging an outage, agents navigating apps and live experiences where every second counts.

Codex and ChatGPT Work
In Codex and ChatGPT Work, Ultrafast is available on Pro $500, eligible usage-based Enterprise and credit-based Edu plans, according to OpenAI’s Speed documentation. For Enterprise workspaces it is off by default, and owners can enable it for selected users or the whole workspace.
The same documentation says Ultrafast uses included subscription limits at 8x the Standard rate and bills purchased credits and Enterprise pay-as-you-go usage at 6x. Fast mode uses 2.5x and 2x. Other self-serve plans do not have access at launch, even with purchased credits.

In the API
OpenAI’s API guide says Ultrafast for GPT-6 Astra and GPT-6.1 Sol is available to all API users, with separate rate limits from Standard and Fast. For GPT-6.1 Sol the default Ultrafast limits are 1 million tokens per minute at the Build tier, 4 million at Launch and 40 million at Grow, against 500,000, 1 million and 5 million for GPT-6 Astra.
The guide strongly recommends WebSockets for agentic applications that make many tool calls in quick succession, because network overhead can reduce the latency gains. GPT-6.1 Sol Ultrafast supports US and EU data residency; GPT-6 Astra Ultrafast supports US residency only.
The Structural Read
Speed has its own price list. OpenAI’s pricing page lists GPT-6.1 Sol at $1 and $5 per million tokens in Batch and Flex, $2 and $10 at Standard, $4 and $20 at Fast and $12 and $60 at Ultrafast.
Subscriptions pay more than credits. In Codex and ChatGPT Work, Ultrafast uses included subscription limits at 8x the Standard rate but bills purchased credits at 6x, according to OpenAI’s Speed documentation.
Capacity is metered separately. Ultrafast has its own rate limits, and OpenAI’s defaults for GPT-6.1 Sol are higher than for GPT-6 Astra at every tier.
OpenAI Developer Community announcement, 8 October 2026
“Built for work where speed and intelligence make a difference: debugging an outage, agents navigating apps, and live experiences where every second counts.”
Three Implications
A 6X STEP ON OUR ARITHMETIC Ultrafast is $12 and $60 per million tokens against $2 and $10 at Standard for GPT-6.1 Sol.
PLAN-GATED IN CODEX Ultrafast in Codex and ChatGPT Work needs Pro $500, eligible Enterprise or credit-based Edu, per OpenAI.
BUILT FOR LIVE WORK OpenAI names outage debugging, agents navigating apps and live experiences as the uses.
The Business Engineer Lens
This story maps onto the Business Engineer framework Tokenomics: The Economics of AI.
The framework’s starting point: the token “is simultaneously the unit of cognition the model produces, the unit of compute the data center serves, the unit of price the lab charges, and the unit of value the enterprise extracts.”
As we read it, Ultrafast adds speed as a separate dimension of the unit of price: the same GPT-6.1 Sol token now sells at several prices depending on how quickly it is served.
What Is Not Established
We read OpenAI’s announcement, its API guide, the Codex Speed documentation, the API pricing page and the ChatGPT rate card. The speed figure is OpenAI’s own “up to” claim, and we did not test it. We did not contact OpenAI.
The Bottom Line
OpenAI now sells GPT-6.1 Sol in an Ultrafast mode at $12 and $60 per million tokens, six times its Standard API price on our arithmetic, and in Codex and ChatGPT Work it draws included usage at 8x and paid credits at 6x the Standard rate.
94,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. We read OpenAI’s announcement of 8 October 2026, its Ultrafast API guide, the Codex Speed documentation, the API pricing page and the ChatGPT rate card; speed figures are OpenAI’s own and the 6x price ratio is our arithmetic. We did not contact OpenAI. Nothing here is a forecast, and nothing here is financial or investment advice.
Sources: OpenAI Developer Community: Ultrafast is rolling out today for GPT-6.1 Sol in the API, Codex, and ChatGPT Work (8 Oct 2026) · OpenAI API docs: Ultrafast mode · OpenAI Codex docs: Speed · OpenAI API pricing









