Fast mode for API Customers

Fast mode offers reliable, high-speed performance with the flexibility to pay-as-you-go. For our latest frontier model, gpt-5.6-sol, you can access up to 2.5x faster speeds with Fast mode.

By choosing Fast mode, you can unlock:

Predictably low latency: Fast mode generates tokens faster and at a more consistent speed than the Standard processing service, even during peak demand.
Easy-to-use flexibility: Like Standard processing, Fast mode can be accessed on a flexible, pay-as-you-go basis instead of requiring advance provisioning.

Note: Priority processing was renamed Fast mode on July 30, 2026. You can use either service_tier: priority or service_tier: fast in your API requests.

	Price per 1M input tokens	Price per 1M input tokens (cached)	Price per 1M output tokens	Uptime SLA³	Latency SLA³
GPT-5.6 Sol excludes long context¹	$10.00	$1.00	$60.00	99.9%	99% > 80 tokens per second²
GPT-5.6 Terra excludes long context¹	$4.00	$0.40	$24.00	99.9%	99% > 70 tokens per second²
GPT-5.6 Luna excludes long context¹	$0.40	$0.04	$2.40	99.9%	99% > 100 tokens per second²
GPT-5.5 excludes long context¹	$12.50	$1.250	$75.00	99.9%	99% > 50 tokens per second²
GPT-5.4 mini excludes long context¹	$1.50	$0.150	$9.00	99.9%	99% > 100 tokens per second²
GPT-5.4 excludes long context¹	$5.00	$0.500	$30.00	99.9%	99% > 50 tokens per second²
GPT-5.2	$3.50	$0.350	$28.00	99.9%	99% > 50 tokens per second²
GPT-5.1	$2.50	$0.250	$20.00	99.9%	99% > 50 tokens per second²
GPT-5	$2.50	$0.250	$20.00	99.9%	99% > 50 tokens per second²
GPT-5 mini	$0.45	$0.045	$3.60	99.9%	99% > 80 tokens per second²
GPT-5.1 codex	$2.50	$0.250	$20.00	99.9%	99% > 50 tokens per second²
GPT-5 codex	$2.50	$0.250	$20.00	99.9%	99% > 50 tokens per second²
GPT-4.1	$3.50	$0.875	$14.00	99.9%	99% > 80 tokens per second²
GPT-4.1 mini	$0.70	$0.175	$2.80	99.9%	99% > 90 tokens per second²
GPT-4.1 nano	$0.20	$0.050	$0.80	99.9%	99% > 100 tokens per second²
GPT-4o gpt-4o-2024-11-20 gpt-4o-2024-08-06	$4.25	$2.125	$17.00	99.9%	99% > 80 tokens per second²
gpt-4o-2024-05-13	$8.75	—	$26.25	99.9%	99% > 80 tokens per second²
GPT-4o mini	$0.25	$0.125	$1.00	99.9%	99% > 90 tokens per second²
o3	$3.50	$0.875	$14.00	99.9%	99% > 80 tokens per second²
o4-mini	$2.00	$0.500	$8.00	99.9%	99% > 90 tokens per second²

1Requests estimated at >272K prompt tokens

2Calculated as p50 request latency on a per 5 minute basis. For customers with existing enterprise agreements that have latency SLAs calculated as p50 request latency on a per minute basis, the prior SLAs are also still applicable.

3This is applicable to Enterprise customers only

How it works

Customers can direct traffic to Fast mode on a per request basis using the existing service_tier parameter, with the option service_tier = "fast".

Tokens served by Fast mode will be billed on a per-token basis, priced at a premium relative to Standard processing rates.

In addition to being configured at the request level, you can also default a project to Fast mode in Project settings > Default Service Tier: Fast. You can still override per request. Selecting Fast in your project settings is equivalent to selecting Priority.

Limitations

Fast mode rate limits are shared with other service tiers.
In rare cases, rapid increases to your Fast mode Tokens per Minute can lead to hitting ramp rate limits. If you exceed the ramp rate limit, then additional traffic may be sent to Standard processing instead.

Frequently asked questions

On July 30, 2026, Priority processing was renamed to Fast mode, and we increased the speeds at which we run gpt-5.6-sol to make it 2.5x faster than standard speeds. The name Fast mode more clearly communicate its primary benefit: faster responses for latency-sensitive workloads. The new name reflects the same flexible, pay-as-you-go processing customers already use. This is not a breaking change as both service_tier=priority and service_tier=fast continue to be accepted by our API to access this functionality.

For existing models, you can set service_tier to either priority or fast in requests to /v1/responses, /v1/chat/completions, and /v1/usage. Both options will continue to return priority in API responses and appear as priority in the Usage dashboard and invoices.

For models released after gpt-5.6, API responses, usage reporting, and invoices will reflect the fast mode name.

Pricing

Models

Rate limits

Fast mode has ramp rate limits to ensure consistently high performance for all customers, while still providing flexible, on-demand pricing. If (a) Fast mode performance is degraded AND (b) a customer's traffic is ramping too quickly, then some Fast mode may be downgraded to Standard processing instead.

The current Fast mode ramp rate limit is defined as processing at least 1M TPM, and increasing traffic by >50% Tokens Per Minute in less than 15 minutes.

Requests processed by Standard service tier will be billed at standard rates, and are not eligible for Priority processing Service Level Objectives.

Requests processed by Standard service tier will include service_tier="Default" in the response.

Best practices for staying within your ramp rate limit

Gradually increase traffic when changing models. For example, if your application is transitioning from a previous snapshot to a new one, use a feature flag to transition traffic over the course of a few hours rather than all at once.
Avoid running large data processing or asynchronous jobs on Fast mode. These jobs can ramp traffic very quickly, and often do not need the improved performance of Fast mode.
If you routinely encounter ramp rate limits, consider purchasing Scale Tier capacity instead or in addition.

Reliability

Policies

The current Fast mode ramp rate limit is defined as processing at least 1M TPM, and increasing traffic by >50% Tokens Per Minute in less than 15 minutes.

Requests processed by Standard service tier will be billed at standard rates, and are not eligible for Priority processing Service Level Objectives.

Requests processed by Standard service tier will include service_tier="Default" in the response.

Best practices for staying within your ramp rate limit

Gradually increase traffic when changing models. For example, if your application is transitioning from a previous snapshot to a new one, use a feature flag to transition traffic over the course of a few hours rather than all at once.
Avoid running large data processing or asynchronous jobs on Fast mode. These jobs can ramp traffic very quickly, and often do not need the improved performance of Fast mode.
If you routinely encounter ramp rate limits, consider purchasing Scale Tier capacity instead or in addition.

Fast mode for API Customers

How it works

Limitations

Frequently asked questions

What does the change from Priority processing to Fast mode mean for customers?

Pricing

(For Enterprise customers) How does this interact with Scale Tier?

(For Enterprise customers) Can I automatically send my Scale Tier spill-over traffic to Fast mode?

(For Enterprise customers) Is my annual commitment tied to a specific processing mode?

Do I still get a discount on Cached input tokens?

How do I view my Fast mode usage and spend?

Models

Is Fast mode available for long context, fine-tuned models, embeddings, etc.?

How do other modalities work with Fast mode?

Will future models be supported?