What is Sol Ultrafast list price?

Short context: $12 input, $0.60 cached input, $15 cache write, and $60 output per million tokens. That is 6x Sol Standard ($2 / $0.10 / $2.50 / $10). Long context (>272K): $24 / $1.20 / $30 / $90. Retrieved 2026-10-08 from the OpenAI pricing Ultrafast tab.

How do I call Sol Ultrafast?

Set model to gpt-6.1-sol and service_tier to ultrafast. It is a serving tier on the existing model, not a new model ID. WebSockets are recommended for tool-heavy agent loops.

What are the default Sol Ultrafast TPM limits?

Build 1,000,000 tokens per minute, Launch 4,000,000, and Grow 40,000,000. Ultrafast rate limits are separate from Standard and Fast. Higher limits go through your OpenAI account team.

Is Fast the same as Ultrafast?

No. Fast is 2x Standard ($4 / $0.20 / $5 / $20 short). Ultrafast is 6x Standard ($12 / $0.60 / $15 / $60 short). Keep them on separate Order Form lines.

Does Sol Ultrafast support EU data residency?

Yes. Sol Ultrafast supports US and EU data residency and global processing. Regional processing still adds +10% where used. Astra Ultrafast remains US residency and global only.

Who can use Sol Ultrafast in Codex and ChatGPT Work?

Pro 500, eligible usage-based Enterprise, and credit-based Edu. Enterprise admins must enable access. API Ultrafast is open to all API users. See the Pro 500 seat cite for packaging; do not mash the meters.

Is Sol Ultrafast the same as Astra Ultrafast?

No. Same 6x multiplier pattern, different dollars. Sol Ultrafast short is $12 / $60. Astra Ultrafast short is $60 / $300. Sol Ultrafast is about one-fifth of Astra Ultrafast on input and output. Separate TPM tables and residency rules.

← Reality Check

GPT-6.1 Sol Ultrafast Costs 6x Standard. Which Calls Belong on the Faster Meter?

On Oct 8 2026 OpenAI put Ultrafast on gpt-6.1-sol. Short API list is $12 input / $0.60 cached / $15 cache write / $60 output per million tokens, 6x Sol Standard $2/$0.10/$2.50/$10. Long context (>272K) Ultrafast is $24/$1.20/$30/$90. Vendor claims up to 6x faster token generation in the API and up to 8x in Codex vs Sol Standard (not guaranteed end-to-end). API is open to all users with separate Ultrafast TPM (Build 1M / Launch 4M / Grow 40M). Codex and ChatGPT Work need Pro 500 or eligible Enterprise/Edu. US and EU data residency supported. Our Sep 29 Sol cite said Ultrafast had no Sol row. That line is now wrong. Map the Sol Ultrafast meter before agents default to it.

Jonathan Kvarfordt · Published October 8, 2026 · 7 min read

Why trust this analysis?

The short answer

When Sol Ultrafast costs 6x Standard list, which Order Form lines must we name before agents default to service_tier ultrafast on gpt-6.1-sol?

On Oct 8 2026 OpenAI made Ultrafast broadly available for gpt-6.1-sol in the API. Call it with model gpt-6.1-sol and service_tier ultrafast. Short Ultrafast list is $12 input / $0.60 cached / $15 cache write / $60 output per MTok, 6x Sol Standard $2 / $0.10 / $2.50 / $10. Long context (>272K) Ultrafast is $24 / $1.20 / $30 / $90. Fast remains 2x Standard ($4 / $0.20 / $5 / $20 short). Do not confuse Fast with Ultrafast. Ultrafast TPM is a separate pool: Build 1M, Launch 4M, Grow 40M. Codex and ChatGPT Work Ultrafast need Pro 500 or eligible Enterprise/Edu (admin enable). US and EU data residency are supported; regional processing is +10% where used. Vendor claims up to 6x faster token generation in the API and up to 8x in Codex vs Sol Standard are VENDOR, not end-to-end guarantees. 1M in + 1M out is about $12 on Standard vs about $72 on Ultrafast before other charges. Sol Ultrafast short is about one-fifth of Astra Ultrafast $60 / $300 on input and output. Name default tier, rate card, Ultrafast TPM, chargeback owner, seat gate, regional uplift, and accepted-outcome owner before Ultrafast is the agent default.

Decision rule

Budget the 6x list and the separate Ultrafast TPM pool, not the tok/s headline. Name when ultrafast is allowed, the short and long rates, chargeback, seat gate, regional uplift, and which latency-sensitive calls justify 6x before agents default to it.

Evidence

developers.openai.com/api/docs/pricing (Ultrafast tab, retrieved 2026-10-08): gpt-6.1-sol short Ultrafast $12 / $0.60 / $15 / $60; long Ultrafast $24 / $1.20 / $30 / $90; Standard short $2 / $0.10 / $2.50 / $10; Fast short $4 / $0.20 / $5 / $20; Astra Ultrafast short $60 / $6 / $75 / $300
developers.openai.com/api/docs/models/gpt-6.1-sol and Ultrafast mode guide (retrieved 2026-10-08): model gpt-6.1-sol + service_tier ultrafast; Ultrafast prices are 6x Standard; Sol Ultrafast default TPM Build 1M / Launch 4M / Grow 40M; Ultrafast has separate rate limits from Standard and Fast; US and EU data residency supported; WebSockets recommended for tool loops
OpenAI Community + @OpenAIDevs (Oct 8 2026): Ultrafast rolling out for GPT-6.1 Sol in the API, Codex, and ChatGPT Work; Codex/Work on Pro 500, eligible usage-based Enterprise, and credit-based Edu; Enterprise admins must enable; vendor speed claims up to 6x API / 8x Codex vs Sol Standard

Operator action

Today: write gpt-6.1-sol + service_tier ultrafast on the Order Form only for named latency-sensitive routes, lock short and long Ultrafast rates, record your Ultrafast TPM tier, check Pro 500 or Ent/Edu admin enable for Codex/Work, name the chargeback owner, and measure cost per accepted task before Ultrafast becomes the default.

Last reviewed October 8, 2026

Six times the Sol sticker just got a Sol row.

On Oct 8 2026 OpenAI put Ultrafast on gpt-6.1-sol in the API, Codex, and ChatGPT Work (OpenAI Community; @OpenAIDevs). Our Sep 29 Sol cite said Ultrafast had no Sol row. That line is now wrong.

The argument

How this reality check breaks down

A map of the sections ahead, in the order the case is made. The sourced charts are in the research library.

Someone on your team will flip an agent to it because it feels fast. This decoder covers what that flip costs and which lines belong on paper first.

What landed

You call it with model `gpt-6.1-sol` and `service_tier: "ultrafast"`. It is a serving tier, not a new model ID (GPT-6.1 Sol; Ultrafast mode).

Reality Check

GPT-6.1 Sol Ultrafast Costs 6x Standard. Which Calls Belong on the Faster Meter?

Source-cited charts live in the research library.

API Ultrafast is open to all API users. Codex and ChatGPT Work Ultrafast need Pro 500, eligible usage-based Enterprise, or credit-based Edu, and Enterprise admins must enable access (OpenAI Community). Seat packaging stays on Pro 200 / Pro 500 Ultrafast. Do not rehash that seat cut here.

The list ladder (locked)

Short-context list, $ per million tokens (MTok), from the pricing Ultrafast tab, retrieved 2026-10-08 (OpenAI pricing). Order: input / cached input / cache write / output.

Standard: $2 / $0.10 / $2.50 / $10 (baseline).
Fast: $4 / $0.20 / $5 / $20 (2x Standard). Do not confuse Fast with Ultrafast.
Ultrafast: $12 / $0.60 / $15 / $60 (6x Standard).
Long Ultrafast (>272K input): $24 / $1.20 / $30 / $90.

Sol API serving tiers short-context list per MTok: Standard $2 / $0.10 / $2.50 / $10; Fast 2x at $4 / $0.20 / $5 / $20; Ultrafast 6x at $12 / $0.60 / $15 / $60 with service_tier ultrafast. — gpt-6.1-sol short-context list per MTok. Retrieved 2026-10-08 from OpenAI pricing. Long Ultrafast is $24 / $1.20 / $30 / $90. Speed claim is vendor.

Rough bill check: 1M input + 1M output is about $12 on Standard vs about $72 on Ultrafast before cache, long-context, or regional charges.

Sol Ultrafast short is about one-fifth of Astra Ultrafast $60 / $6 / $75 / $300 on input and output, with the same 6x multiplier pattern against each model's own Standard.

Separate Ultrafast TPM

Ultrafast has separate rate limits from Standard and Fast. Default Sol Ultrafast TPM by API usage tier (Ultrafast mode):

Build: 1,000,000 TPM.
Launch: 4,000,000 TPM.
Grow: 40,000,000 TPM.
Higher limits: request through your OpenAI account team.

A volume agent that hits the Ultrafast ceiling falls back or queues. Know which happens before go-live. Do not assume Standard or Fast headroom covers Ultrafast traffic.

Region and transport

Sol Ultrafast supports US and EU data residency and global processing (Ultrafast mode; model page). That is wider than Astra Ultrafast, which is US residency and global only. Regional processing still adds +10% where used.

OpenAI strongly recommends WebSockets for agentic tool loops. HTTP works, but connection overhead on every tool call can erase the latency you paid 6x to get.

The speed claim

OpenAI claims up to 6x faster token generation in the API and up to 8x in Codex versus Sol Standard. Tag it VENDOR. It measures tokens out per second, not end-to-end task time, and it is not a guarantee. Your agent still spends time in tool calls, retrieval, and retries.

Decision rule

Order Form checklist

Default tier: name when `service_tier: "ultrafast"` is allowed vs Standard and Fast.
Rate card: short Ultrafast $12 / $0.60 / $15 / $60; long Ultrafast $24 / $1.20 / $30 / $90.
Ultrafast TPM: separate pool Build 1M / Launch 4M / Grow 40M; not the Standard+Fast pool.
Chargeback owner: who owns Ultrafast burn versus Standard and Fast.
Codex / Work gate: Pro 500 or eligible Enterprise / Edu; Ent admin enable.
Regional +10%: US and EU residency supported; print the uplift where used.
Accepted-outcome owner: which latency-sensitive calls justify 6x before agents default.
Distinct meters: not Astra Ultrafast rates; not Pro 500 seat Ultrafast packaging.

Keep the meters apart

[GPT-6.1 Sol near-Astra at one-fifth price](/blog/gpt-6-1-sol-near-astra-fifth-price-task-cost-order-form): Sol Standard list and volume default. Ultrafast is now live; see this cite for the Ultrafast meter.
[Astra Ultrafast API 6x](/blog/gpt-6-astra-ultrafast-api-6x-order-form): Astra Ultrafast rates, TPM, and US-only residency. Different model, different dollars.
[Pro 200 / Pro 500 Ultrafast](/blog/openai-pro-200-usage-cut-pro-500-ultrafast-order-form): ChatGPT seat packaging for Codex/Work Ultrafast. Cross-link only.

Sources

OpenAI API pricing, Ultrafast tab: developers.openai.com/api/docs/pricing
OpenAI, GPT-6.1 Sol model docs: developers.openai.com/api/docs/models/gpt-6.1-sol
OpenAI, Ultrafast mode guide: developers.openai.com/api/docs/guides/ultrafast-mode
OpenAI Community, Ultrafast for GPT-6.1 Sol (Oct 8 2026): community.openai.com
@OpenAIDevs (Oct 8 2026): status/2108262812489531498, status/2108262830348886150

If you want the next Order Form decoder before agents burn the 6x Sol tier, subscribe.

Take it to the room

The short list from this piece

Pulled from the argument above, written so you can read it out in a pipeline or board review.

Frequently asked questions

What is Sol Ultrafast list price?: Short context: $12 input, $0.60 cached input, $15 cache write, and $60 output per million tokens. That is 6x Sol Standard ($2 / $0.10 / $2.50 / $10). Long context (>272K): $24 / $1.20 / $30 / $90. Retrieved 2026-10-08 from the OpenAI pricing Ultrafast tab.
How do I call Sol Ultrafast?: Set model to gpt-6.1-sol and service_tier to ultrafast. It is a serving tier on the existing model, not a new model ID. WebSockets are recommended for tool-heavy agent loops.
What are the default Sol Ultrafast TPM limits?: Build 1,000,000 tokens per minute, Launch 4,000,000, and Grow 40,000,000. Ultrafast rate limits are separate from Standard and Fast. Higher limits go through your OpenAI account team.
Is Fast the same as Ultrafast?: No. Fast is 2x Standard ($4 / $0.20 / $5 / $20 short). Ultrafast is 6x Standard ($12 / $0.60 / $15 / $60 short). Keep them on separate Order Form lines.
Does Sol Ultrafast support EU data residency?: Yes. Sol Ultrafast supports US and EU data residency and global processing. Regional processing still adds +10% where used. Astra Ultrafast remains US residency and global only.
Who can use Sol Ultrafast in Codex and ChatGPT Work?: Pro 500, eligible usage-based Enterprise, and credit-based Edu. Enterprise admins must enable access. API Ultrafast is open to all API users. See the Pro 500 seat cite for packaging; do not mash the meters.
Is Sol Ultrafast the same as Astra Ultrafast?: No. Same 6x multiplier pattern, different dollars. Sol Ultrafast short is $12 / $60. Astra Ultrafast short is $60 / $300. Sol Ultrafast is about one-fifth of Astra Ultrafast on input and output. Separate TPM tables and residency rules.

Share this page

Posting to Instagram or TikTok? Copy the link, it carries the title, summary and share image.

Get the next Reality Check before you sign the order form.

Keep reading

Reality Check

AI Review Time Is a Line Item. Price It and Name the Owner.

On TextQL's Argo-Bench the best AI model fully solved 34.8% of 210 data tasks. Put review time and a named owner into your AI business case.

Reality Check

Call Center Headcount Is Falling. Can You Prove AI Did It?

Revelio Labs says global call-center headcount is 5.3% below its Dec 2023 peak. Revelio: "The decline is surely multifaceted." How to test your own org.