Is Grok 4 Fast available through the xAI API, and what does it cost?
Grok 4 Fast is no longer served as a separate xAI API model. xAI’s retirement notice says that, effective May 15, 2026 at 12:00 p.m. PT, the API retired the model IDs grok-4-fast-reasoning and grok-4-fast-non-reasoning. Requests to those IDs still resolve, but are automatically redirected to Grok 4.3—and billed at Grok 4.3 rates, not the retired model’s original rates. (xAI retirement notice)
So, for an existing integration that still sends a retired Grok 4 Fast ID, the documented current budget reference is Grok 4.3 pricing. The old ID resolving successfully does not mean the original Grok 4 Fast model is still available.
Rates for requests redirected to Grok 4.3
The xAI pricing documentation lists these global Grok 4.3 text rates in USD per 1 million tokens:
| Prompt context | Input | Cached input | Output |
|---|---|---|---|
| Short context | $1.25 | $0.20 | $2.50 |
| Long context (prompt reaches 200,000 tokens) | $2.50 | $0.40 | $5.00 |
When the prompt reaches the long-context threshold, the long-context rates apply to all tokens in that request. (xAI API pricing; Grok 4.3 model page)
Example budget
For 1 million uncached input tokens plus 1 million output tokens, the token-cost arithmetic is:
- $3.75 at short-context rates: $1.25 input + $2.50 output.
- $7.50 at long-context rates: $2.50 input + $5.00 output.
These are illustrative estimates based on the listed Grok 4.3 rates. They exclude any applicable tools or other charges. The context tier depends on prompt length, not the sum of input and output tokens.
What happens to reasoning and non-reasoning requests?
The retirement notice says requests using the retired reasoning IDs are redirected to Grok 4.3 with low reasoning effort. Requests using the retired non-reasoning IDs are redirected with no reasoning effort. If you want to control the replacement model or reasoning effort explicitly, update the model choice in your integration rather than relying on the legacy-ID redirect. (xAI retirement notice)
Don’t confuse Grok 4 Fast with Grok 4.7 Fast
The names refer to different documented offerings. The retirement notice identifies the retired API IDs as grok-4-fast-reasoning and grok-4-fast-non-reasoning. Separately, the pricing page describes Grok 4.7 Fast as available only in Cursor and Grok Build, billed through those plans, and unavailable on the public xAI API. Its plan-related rates should not be presented as the current public API price for retired Grok 4 Fast IDs. (xAI retirement notice; xAI API pricing)
For standard Grok 4.7 model information, xAI’s models page recommends it for code and chat and lists a 500,000-token context window and configurable reasoning. Those details describe Grok 4.7, not the retired Grok 4 Fast API model.
Bottom line
For the old Grok 4 Fast API IDs, xAI documents automatic redirection to Grok 4.3 and billing at Grok 4.3 rates. Budget using the applicable Grok 4.3 token rates above; don’t use Grok 4.7 Fast plan rates or assume the retired model’s original pricing still applies.
Sources
May 15, 2026 Model Retirement | xAI Documentation