# OpenAI Ultrafast vs Sol: $7.20 vs $0.22 for the Same Request

> Ultrafast is an OpenAI service tier, not a model. As of October 3, 2026, only GPT-6 Astra has an Ultrafast price; GPT-6.1 Sol runs on Standard and Fast.

- Source: https://www.aifreeapi.com/en/posts/openai-ultrafast-vs-sol
- Language: en
- Published: 2026-10-03
- Updated: 2026-10-03
- Publisher: AI Free API (https://www.aifreeapi.com)

Ultrafast and Sol are not two options on the same menu. Ultrafast is a service tier of the OpenAI API, alongside Standard, Batch, Flex and Fast. Sol is a model family: GPT-6.1 Sol, GPT-6 Sol and GPT-5.6 Sol. The real choice is which model you run on which tier, and as of October 3, 2026, that choice is narrower than the September 29 announcement suggested:

- **GPT-6 Astra** is the only model every API account can run on Ultrafast. It costs $60 input and $300 output per 1M tokens, which is 6× its Standard rate.
- **GPT-6.1 Sol** runs on Standard ($2 input, $10 output) and Fast ($4 input, $20 output). OpenAI announced GPT-6.1 Sol Ultrafast, but its documentation lists no price for it and no way to call it.
- **GPT-5.6 Sol** has Ultrafast only as a preview that you request through an OpenAI account team. It has no public price.

Take one request with 50,000 fresh input tokens, 200,000 cached input tokens and 10,000 output tokens. At list price it costs $0.22 on GPT-6.1 Sol Standard, $0.44 on GPT-6.1 Sol Fast and $7.20 on GPT-6 Astra Ultrafast. So "Ultrafast or Sol" today means "pay about 33 times more for Astra at top speed, or stay on Sol." The sections below show who has access, how the rates are built, and when the premium is worth paying.

## Ultrafast vs Sol: a service tier and a model family

Ultrafast is how a request is served; Sol is which model answers it. OpenAI's [Ultrafast mode guide](https://developers.openai.com/api/docs/guides/ultrafast-mode) describes it as "the fastest service tier in the OpenAI API" and adds: "Use it when speed justifies the higher cost." You select it with a request parameter, `service_tier`, not with a model name.

The confusion has a history. In August 2026, OpenAI [previewed Ultrafast](https://openai.com/index/previewing-ultrafast/) with GPT-5.6 Sol as its only model, so "GPT-5.6 Sol Ultrafast" spread as if it were a product name. On September 29, OpenAI released GPT-6.1 Sol (API ID `gpt-6.1-sol`) and, in the [same post](https://openai.com/index/introducing-gpt-6-1-sol/), announced GPT-6 Astra Ultrafast and GPT-6.1 Sol Ultrafast. Two things launched together, and one of them carries the name of both.

"Fast" is a separate, cheaper tier. It is the current name of what OpenAI called Priority processing until July 30, 2026, and it costs 2× Standard.

## Which Sol models can run on Ultrafast as of October 3, 2026

No Sol model is generally available on Ultrafast. GPT-6 Astra is the only model OpenAI documents as open to all API users on that tier.

| Model | Standard and Fast tiers | Ultrafast in the API | Ultrafast in Codex and ChatGPT Work |
| --- | --- | --- | --- |
| GPT-6 Astra | Yes | All API users, at low default rate limits | Pro $500 and eligible Enterprise and Edu plans |
| GPT-6.1 Sol | Yes | Announced September 29; not in the Ultrafast guide or pricing table | Not listed; the docs say it "supports Standard and Fast where available" |
| GPT-5.6 Sol | Standard at a promotional price; Fast in Codex | Preview only, requested through an OpenAI account team; no public price | Not listed |

The GPT-6.1 Sol row needs its date. OpenAI's announcement names GPT-6.1 Sol Ultrafast as offered through the API and to subscribers of the new $500-per-month Pro tier. [VentureBeat's launch report](https://venturebeat.com/technology/openais-gpt-6-1-sol-offers-astra-like-performance-at-1-5th-price-a-new-ultrafast-tier-clocks-at-300-tokens-per-second) says the Sol version is "due in the coming days." The pages you would actually build against agree with VentureBeat. The Ultrafast guide names only GPT-6 Astra and GPT-5.6 Sol. The Ultrafast tab of the pricing table has one row. The [Codex speed page](https://learn.chatgpt.com/docs/agent-configuration/speed) lists GPT-6.1 Sol under Standard and Fast only. Announced, then, but neither callable nor priced in the documentation as of October 3, 2026.

To tell when that changes, look for a `gpt-6.1-sol` row on the Ultrafast tab of the [API pricing table](https://developers.openai.com/api/docs/pricing) and for the model name in the guide's availability section.

## OpenAI Ultrafast pricing: $60/$300 for Astra, no Sol row yet

Ultrafast costs 6× the Standard rate of the same model, and the only published Ultrafast rate belongs to GPT-6 Astra. These are the API list prices per 1M tokens for requests with up to 272,000 input tokens, as of October 3, 2026:

| Model and tier | Input | Cached input | Cache writes | Output |
| --- | --- | --- | --- | --- |
| GPT-6.1 Sol, Standard | $2.00 | $0.10 | $2.50 | $10.00 |
| GPT-6.1 Sol, Fast | $4.00 | $0.20 | $5.00 | $20.00 |
| GPT-5.6 Sol, Standard (promotional) | $4.00 | $0.40 | $5.00 | $20.00 |
| GPT-6 Astra, Standard | $10.00 | $1.00 | $12.50 | $50.00 |
| GPT-6 Astra, Fast | $20.00 | $2.00 | $25.00 | $100.00 |
| GPT-6 Astra, Ultrafast | $60.00 | $6.00 | $75.00 | $300.00 |
| GPT-6.1 Sol, Ultrafast | Not published | Not published | Not published | Not published |

Three conditions change these numbers. Above 272,000 input tokens, long-context rates apply: GPT-6 Astra Ultrafast becomes $120 input and $450 output, and GPT-6.1 Sol Standard becomes $4 input and $15 output. Regional processing endpoints add 10%, although Ultrafast does not run on EU or other non-US regional endpoints at all. And the GPT-5.6 Sol rate is promotional; the pricing page guarantees it only "at least through November 21, 2026."

You will see $12 input, $0.60 cached input and $60 output quoted for GPT-6.1 Sol Ultrafast. That is VentureBeat's arithmetic, 6× the Standard rate, and VentureBeat labels it as derived. OpenAI has published no such price. Its own statement is looser: with GPT-6.1 Sol Ultrafast, developers would pay roughly what GPT-6 Astra used to cost. Astra's Standard rate is $10 input and $50 output, so the two are consistent, but only one of them is a price list.

## Cost of one request: $0.22 on GPT-6.1 Sol, $7.20 on Astra Ultrafast

The same request costs $0.22 on GPT-6.1 Sol Standard and $7.20 on GPT-6 Astra Ultrafast, a 32.7× difference. The request is the one from the opening: 50,000 uncached input tokens, 200,000 cached input tokens and 10,000 output tokens. That is 250,000 input tokens in total, inside the short-context band. Cache-write charges are left out.

The formula is (0.05 × input rate) + (0.2 × cached input rate) + (0.01 × output rate):

| Model and tier | Calculation | Cost |
| --- | --- | --- |
| GPT-6.1 Sol, Standard | 0.05 × $2 + 0.2 × $0.10 + 0.01 × $10 | $0.22 |
| GPT-6.1 Sol, Fast | 0.05 × $4 + 0.2 × $0.20 + 0.01 × $20 | $0.44 |
| GPT-5.6 Sol, Standard | 0.05 × $4 + 0.2 × $0.40 + 0.01 × $20 | $0.48 |
| GPT-6 Astra, Standard | 0.05 × $10 + 0.2 × $1 + 0.01 × $50 | $1.20 |
| GPT-6 Astra, Fast | 0.05 × $20 + 0.2 × $2 + 0.01 × $100 | $2.40 |
| GPT-6 Astra, Ultrafast | 0.05 × $60 + 0.2 × $6 + 0.01 × $300 | $7.20 |
| GPT-6.1 Sol, Ultrafast (hypothetical, 6× Standard) | 6 × $0.22 | $1.32 |

![Bar chart of one identical request priced on six OpenAI model and tier combinations, from $0.22 on GPT-6.1 Sol Standard to $7.20 on GPT-6 Astra Ultrafast](https://www.aifreeapi.com/posts/en/openai-ultrafast-vs-sol/img/ultrafast-vs-sol-request-cost.webp)

Two things are worth reading off this table. First, the tier multiplier is uniform, so your token mix does not change the 6× between Astra Standard and Astra Ultrafast. The gap between models does move with the mix: Astra's input and output rates are 5× GPT-6.1 Sol's, but its cached input rate is 10× ($1.00 against $0.10). The more of your prompt is cached, the wider the gap. Second, the hypothetical $1.32 for GPT-6.1 Sol Ultrafast lands next to Astra Standard's $1.20. That is the trade OpenAI described, and it does not exist as a price yet.

This compares list price for an identical request. Different models spend different numbers of tokens on the same job, so it is not a cost per completed task. For how far those two measures can drift apart, see [GPT-6.1 Sol vs Claude Opus 5.5: Half the Price, 1/6 the Task Cost](/en/posts/gpt-6-1-sol-vs-claude-opus-5-5).

## How fast Ultrafast is: up to 6× in the API, up to 8× in Codex

Every published speed figure is an "up to" maximum for generating output tokens, and none of them describes how long your whole task takes.

| Figure | Applies to | Source |
| --- | --- | --- |
| Up to 8× GPT-6 Astra in Standard mode | GPT-6 Astra Ultrafast in Codex | OpenAI's Codex speed page |
| Up to 6× in the API, up to 300 tokens per second | GPT-6 Ultrafast | VentureBeat's account of OpenAI's DevDay materials |
| Up to 14× Standard, up to 750 output tokens per second | GPT-5.6 Sol Ultrafast preview, powered by Cerebras | OpenAI's August 13 preview post |

OpenAI is explicit about the scope of its own number: "This comparison measures token generation speed, not billing rates or overall task completion time." Network round trips, tool execution and input processing sit outside the multiplier. OpenAI publishes no Standard-mode tokens-per-second figure for GPT-6 Astra or GPT-6.1 Sol, and its pricing page gives none for Fast mode either. None of the figures above come from an independent benchmark.

A rough sense of scale, using the reported ceiling: 10,000 output tokens at 300 tokens per second take about 33 seconds to generate. At one sixth of that rate, 50 tokens per second, they take about 200 seconds. The 50 is implied by "6× in the API," not published, so treat that gap as a best case.

## Calling Ultrafast in the API: service_tier, WebSockets, rate limits

Set `model` to `gpt-6-astra` and `service_tier` to `ultrafast` on the Responses API. This is the HTTP form from OpenAI's guide:

```bash
curl https://api.openai.com/v1/responses \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6-astra",
    "input": "Explain why the sky is blue in one sentence.",
    "service_tier": "ultrafast"
  }'
```

HTTP works, but OpenAI steers you to a persistent connection. The guide strongly recommends WebSockets, "especially for agentic applications that make many tool calls in quick succession. Without a persistent connection, network overhead can reduce the latency gains." In the Python SDK that means opening one connection with `client.responses.connect()`, sending each turn with `connection.response.create(...)`, and passing the previous response's ID as `previous_response_id` so later turns and tool results reuse the same connection.

Three limits decide whether the tier fits your workload:

- **Rate limits.** The default Ultrafast limits for GPT-6 Astra are 500,000 tokens per minute on API usage tiers 1–3, 1,000,000 on tier 4 and 5,000,000 on tier 5. Higher limits go through an OpenAI account team.
- **Data residency.** "Ultrafast supports US data residency and global processing only. It does not support EU or other non-US regional processing endpoints."
- **Models.** The documented example uses `gpt-6-astra` only. What the API returns when `service_tier: "ultrafast"` is sent with a Sol model on an account without preview access is not documented.

## Ultrafast in Codex: Pro $500 and 8× your included usage

In Codex and ChatGPT Work, Ultrafast is a plan feature, not something you can buy by the token. The Codex speed page says it "is available in Codex and ChatGPT Work on Pro $500 and eligible Enterprise and Edu plans," and that "other self-serve plans don't have access to Ultrafast at launch, even with purchased credits." That excludes Plus and the $100 and $200 Pro plans. In Enterprise workspaces it is off by default, and it is not offered to workspaces that require inference residency outside the United States.

The model is again GPT-6 Astra. GPT-6.1 Sol in Codex has Standard and Fast, which you toggle with `/fast` in the CLI, or persist with `service_tier = "fast"` plus `[features].fast_mode = true` in `config.toml`.

Billing uses two different multipliers, and neither one describes speed:

| Mode in Codex | Draw on included subscription limits | Purchased credits and Enterprise pay-as-you-go |
| --- | --- | --- |
| Fast | 2.5× Standard | 2× Standard |
| GPT-6 Astra Ultrafast | 8× Standard | 6× Standard |

The [Codex pricing page](https://learn.chatgpt.com/docs/pricing) sets Standard credit rates per 1M tokens at 250 input, 25 cached input and 1,250 output for GPT-6 Astra, and 50, 2.5 and 250 for GPT-6.1 Sol. Run the example request through them and GPT-6.1 Sol Standard costs 0.05 × 50 + 0.2 × 2.5 + 0.01 × 250 = 5.5 credits. GPT-6 Astra Standard costs 0.05 × 250 + 0.2 × 25 + 0.01 × 1,250 = 30 credits, and 180 credits on Ultrafast at the 6× purchased-credit rate. That is the same 32.7× as in the API. Included subscription usage drains faster still, at 8×, and OpenAI warns against estimating it from token prices.

If you run Codex with an API key, none of this applies: "Codex uses API token pricing instead, and ChatGPT credit multipliers don't apply." For how Sol's credit billing compares with the cheaper Luna model, see [GPT-6 Luna vs. Sol pricing: calculate the cost of your workload](/en/posts/gpt-6-luna-vs-sol-price), which covers GPT-6 Sol, the predecessor of GPT-6.1 Sol.

## Which to use now: Sol on Fast, Astra on Ultrafast, or wait

Start from GPT-6.1 Sol on Standard and move up only when a specific cost of waiting justifies it.

![Decision flow for choosing between GPT-6.1 Sol Standard, GPT-6.1 Sol Fast, GPT-6 Astra Ultrafast, or waiting for GPT-6.1 Sol Ultrafast](https://www.aifreeapi.com/posts/en/openai-ultrafast-vs-sol/img/ultrafast-or-sol-decision.webp)

- **Nobody is waiting on the output: GPT-6.1 Sol Standard.** Batch jobs, background agents and overnight runs gain nothing from generation speed. At $0.22 for the example request, this is the baseline everything else is measured against.
- **Latency matters and Sol's quality is enough: GPT-6.1 Sol Fast.** It doubles the bill to $0.44, which is still about a third of Astra Standard's $1.20. No tokens-per-second figure is published for it, so time your own requests on both tiers before committing.
- **Someone is waiting on long Astra output: GPT-6 Astra Ultrafast.** In the example it costs $6.00 more than Astra Standard at list price ($7.20 against $1.20). What that buys in time is not published as a figure you can budget against, so the premium makes sense only where the wait itself is the cost. Incident response, customer support, voice and financial research are among the uses OpenAI named for the tier, and each has a person or a deadline on the other end.
- **You want Sol at Ultrafast speed: wait.** GPT-6.1 Sol Ultrafast would be the middle option, near Astra Standard's price if the 6× multiplier holds. As of October 3, 2026, it cannot be called or budgeted. Build on GPT-6.1 Sol Fast now. The tier is a request parameter, so moving up later should not mean restructuring your code.

Three conditions rule Ultrafast out regardless of budget. You need EU or other non-US regional processing. Your account is on a Codex plan below Pro $500. Or your agent spends most of its time in tool calls and network round trips, where faster token generation moves the total very little.

## Questions about Sol, Astra and Ultrafast

### Is GPT-6 Astra better than GPT-6.1 Sol?

By OpenAI's own description, yes on capability: it calls GPT-6 Astra its most capable frontier model overall and positions GPT-6.1 Sol as near-Astra intelligence. Sol wins on price. Its Standard rate is one fifth of Astra's for input and output, and one tenth for cached input.

### Which OpenAI model is the fastest?

By published maximums, GPT-5.6 Sol on Ultrafast, at up to 750 output tokens per second, but it is a limited preview. The fastest model every API account can call is GPT-6 Astra on Ultrafast, reported by VentureBeat at up to 300 tokens per second.

### How much does GPT-6.1 Sol Ultrafast cost?

OpenAI has not published a price as of October 3, 2026. If it follows the 6× multiplier used for GPT-6 Astra, it would be $12 input, $0.60 cached input and $60 output per 1M tokens. That figure is VentureBeat's estimate, not a list price.

### Can I use GPT-5.6 Sol Ultrafast in the API?

Only with preview access. OpenAI's guide tells organizations that work with an OpenAI account team to contact it "to request higher rate limits or preview access for GPT-5.6 Sol." There is no self-serve sign-up and no public price.

### Is OpenAI Ultrafast available on ChatGPT Plus?

No. In Codex and ChatGPT Work, Ultrafast requires Pro $500 or an eligible Enterprise or Edu plan, and buying credits on another self-serve plan does not unlock it. Plus subscribers can use GPT-6.1 Sol on Standard, and on Fast where it is available.