OpenAI Ultrafast pricing: GPT-6.2 API rates and Codex usage
Updated 2026-11-18. From OpenAI's API docs, Codex pricing page and Help Center.
OpenAI Ultrafast costs 3x the Standard rate for GPT-6.2 Astra and GPT-6.2 Sol until Feb 28, 2027, then 6x from March 1. That's $6 / $30 per 1M input and output tokens on GPT-6.2 Sol now. GPT-6 Astra and GPT-6.1 Sol Ultrafast already cost 6x. Fast mode is 2x Standard. On personal ChatGPT plans, only Pro 500 (all four models) and Pro 200 (GPT-6.2 Sol in Codex) get it.
OpenAI Ultrafast pricing for GPT-6.2 Sol is $6 / $30 per million input and output tokens right now, three times the Standard rate. On March 1, 2027 (Pacific time) that doubles to $12 / $60, six times Standard, per OpenAI's API pricing page. GPT-6.2 Astra follows the same rule: $24 / $120 now, $48 / $240 from March.
That promo is the news. OpenAI launched both GPT-6.2 models on Nov 17, 2026, and the older Ultrafast models (GPT-6 Astra and GPT-6.1 Sol) already cost 6x. Every number below is confirmed on OpenAI's own pages, read Nov 18, 2026, except the per-request math, which is our estimate and labeled that way.
Ultrafast vs Fast vs Standard: API price per 1M tokens
Input / output, for prompts up to 272K tokens. Fast mode is simply twice Standard on all four models, so GPT-6.2 Sol on Fast is $4 / $20.
| Model | Standard | Ultrafast now | From Mar 2027 |
|---|---|---|---|
| GPT-6.2 Sol | $2 / $10 | $6 / $30 | $12 / $60 |
| GPT-6.2 Astra | $8 / $40 | $24 / $120 | $48 / $240 |
| GPT-6.1 Sol | $2 / $10 | $12 / $60 | Same |
| GPT-6 Astra | $10 / $50 | $60 / $300 | Same |
Cached input scales the same way: GPT-6.2 Sol cached input is $0.10 per million at Standard and $0.30 on Ultrafast until the promo ends. Prompts over 272K tokens cost more at every tier; our 1M context API pricing guide has the long-prompt rates. Flex pricing isn't listed for the GPT-6.2 models.
What one request costs at each speed
Our estimate for a single agent turn that sends 50,000 input tokens with no cache hits and gets 5,000 tokens back, priced at OpenAI's official rates. Real requests vary, and cache hits make every row cheaper.
| Speed tier | GPT-6.2 Sol | GPT-6.2 Astra |
|---|---|---|
| Standard | $0.15 | $0.60 |
| Batch | $0.075 | $0.30 |
| Fast | $0.30 | $1.20 |
| Ultrafast until Feb 28, 2027 | $0.45 | $1.80 |
| Ultrafast from Mar 1, 2027 | $0.90 | $3.60 |
So the same GPT-6.2 Sol turn costs 3x more on Ultrafast today and 6x more from March. A thousand of these turns a day is $150 at Standard, $450 on Ultrafast now and $900 after the promo. Batch, at half price, is the cheap end when you can wait.
How fast is Ultrafast?
OpenAI calls Ultrafast its "fastest service tier" but doesn't publish a speed multiple for it. For Fast mode, its Fast mode guide says "up to 2.5× faster speeds" than Standard. Fast is the old Priority processing, renamed on July 30, 2026, and the API still accepts service_tier "priority".
To use Ultrafast in the API, set service_tier to "ultrafast". OpenAI's Ultrafast guide recommends WebSockets for agents that make many tool calls, because per-request network overhead eats into the gain. Ultrafast has its own rate limits, separate from Standard and Fast. The published defaults for GPT-6.1 Sol are 1M, 4M and 40M tokens per minute on the Build, Launch and Grow tiers; OpenAI tells you to check your organization's limits for GPT-6.2. GPT-6.2 Ultrafast data residency is US only.
Does Ultrafast use more of your ChatGPT or Codex limit?
Yes, a lot more. Inside ChatGPT Work and Codex, speed modes multiply how fast your included allowance drains (the Ultrafast column), and what purchased credits cost, per OpenAI's Codex pricing page:
| Model | Fast | Ultrafast | Ultrafast on credits |
|---|---|---|---|
| GPT-6.2 Astra or Sol | 2x | 6x | 3x, then 6x from Mar 2027 |
| GPT-6.1 Sol | 2.5x | 8x | 6x |
| GPT-6 Astra | 2.5x | 8x | 6x |
| GPT-6.1 Terra, GPT-6 Sol, GPT-6 Luna | 2.5x | None | None |
Here GPT-6.2 is the better deal: its Ultrafast drains included usage at 6x instead of 8x, and purchased credits cost 3x instead of 6x until Feb 28, 2027. Fast on GPT-6.2 is also a little cheaper, at 2x instead of 2.5x. A heavy Ultrafast session on GPT-6.2 Sol uses the same allowance as six Standard ones, so a weekly limit can go in a day or two.
Which plans get Ultrafast
Pro 200, Pro 500 and eligible Enterprise and Edu workspaces. OpenAI's Work and Codex help article lists:
| ChatGPT plan | Ultrafast access in Work and Codex |
|---|---|
| Plus, Pro 100, Business | None, not even with credits |
| Pro 200 | GPT-6.2 Sol in Codex only |
| Pro 500 | GPT-6.2 Astra, GPT-6.2 Sol, GPT-6 Astra, GPT-6.1 Sol |
| Enterprise, Edu (eligible) | Paid from workspace credits, if the admin allows it |
Pro 200 gained GPT-6.2 Sol Ultrafast in Codex with the Nov 17 launch; before that, Ultrafast was a Pro 500 feature only. In the API, Ultrafast for GPT-6 Astra and GPT-6.1 Sol is open to all API users.
Is Ultrafast worth the price?
Only when a person is waiting on the answer. For live coding in Codex or a user-facing agent where latency is the product, GPT-6.2 Sol at 3x is the cheapest Ultrafast OpenAI sells right now: half the price of GPT-6.1 Sol Ultrafast, which has the same Standard rate.
For background jobs, evals and anything that can finish later, stay on Standard or use Batch. And if you build on the promo, budget for the March 1, 2027 doubling now: your Ultrafast bill for the same traffic will be twice as big.
Sources
- OpenAI API pricing: Standard, Batch, Fast and Ultrafast tables (read Nov 18, 2026)
- OpenAI API docs: Ultrafast mode (read Nov 18, 2026)
- OpenAI API docs: Fast mode (read Nov 18, 2026)
- OpenAI Codex pricing: speed multipliers (read Nov 18, 2026)
- OpenAI Help Center: ChatGPT Work and Codex, Ultrafast section (read Nov 18, 2026)
FAQ
How much does GPT-6.2 Ultrafast cost?
GPT-6.2 Sol Ultrafast is $6 / $30 per million input and output tokens through Feb 28, 2027, then $12 / $60. GPT-6.2 Astra is $24 / $120, then $48 / $240, per OpenAI's API pricing page.
When does the Ultrafast promo price end?
February 28, 2027. From March 1, 2027 (Pacific time), Ultrafast for GPT-6.2 Astra and GPT-6.2 Sol costs 6 times the Standard rate instead of 3 times.
Is Ultrafast the same as Fast mode?
No. Fast is the renamed Priority processing at 2x the Standard API price and up to 2.5x the speed. Ultrafast is a separate, faster tier with its own rate limits, sold on four models only.
Does ChatGPT Pro 200 include Ultrafast?
Partly. Since Nov 17, 2026, Pro 200 includes GPT-6.2 Sol Ultrafast in Codex only. Pro 500 gets Ultrafast on GPT-6.2 Astra, GPT-6.2 Sol, GPT-6 Astra and GPT-6.1 Sol.
How much faster does Ultrafast use my Codex limit?
6x the Standard rate on GPT-6.2 models and 8x on GPT-6 Astra and GPT-6.1 Sol, per OpenAI's Codex pricing page. Purchased credits cost 3x on GPT-6.2 until Feb 28, 2027.
Can I use Ultrafast on ChatGPT Plus?
No. OpenAI says Ultrafast is unavailable on Plus, Pro 100 and Business, including through purchased credits.
Related
- AI model API prices compared
- GPT-6.2 Sol API pricing
- GPT-6.2 Astra API pricing
- GPT-6.2 Astra vs GPT-6.2 Sol
- Codex credits: prices and rates
- Codex usage limits explained
- ChatGPT Pro 100 vs 200 vs 500
- Codex pricing and changes
- Claude Fable 5.2 fast mode pricing
More guides
- OpenAI vs Anthropic API pricing
- OpenAI vs Gemini API pricing
- GLM API pricing explained
- Kimi API pricing explained
- MiniMax API pricing explained
- Qwen API pricing tiers explained
Not sure a plan beats the API for you? Try the plan vs API cost calculator.