Grok 4 Fast is no longer available as a distinct xAI API model. xAI lists both grok-4-fast-reasoning and grok-4-fast-non-reasoning among the IDs retired on May 15, 2026, at 12:00 p.m. PT. Requests to those IDs are redirected to grok-4.3; xAI recommends explicitly choosing the replacement model. (docs.x.ai)
For a new integration, use grok-4.3 directly. That makes the model selection and its billing explicit instead of relying on an old Grok 4 Fast ID that now redirects.
Is Grok 4 Fast still available through the API?
No—not as a separately callable model. xAI's retirement notice includes both Grok 4 Fast variants. Requests using grok-4-fast-reasoning are served by grok-4.3 with low reasoning effort; requests using grok-4-fast-non-reasoning are served with no reasoning effort. The notice also says requests to these retired IDs are billed at Grok 4.3 rates. (docs.x.ai)
Use grok-4.3 for a new request. An older Cursor walkthrough uses grok-4-fast-reasoning, but that historical example predates the retirement; it is not a current model ID to build around.
Make a minimal request to Grok 4.3
Set your xAI API key as an environment variable, such as XAI_API_KEY, rather than putting the secret directly in source code. The xAI API reference specifies bearer-token authentication; the Responses API accepts a text input and uses POST /v1/responses. (docs.x.ai)
For example, from a shell where XAI_API_KEY is set:
curl https://api.x.ai/v1/responses \
-H "Authorization: Bearer $XAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "grok-4.3",
"input": "Reply with a short greeting."
}'
The request uses the documented Responses route, bearer-token header, and string input. The retirement notice identifies grok-4.3 as the replacement for both retired Grok 4 Fast IDs. (docs.x.ai)
What success looks like
Check the API response body for a model value of grok-4.3, a status of completed, and an output array containing the generated response. The Responses reference defines these fields and lists completed, in_progress, and incomplete as possible status values. (docs.x.ai)
A model answering “What model are you?” is not proof of which model handled a request. Use the API response fields as the check; if the request fails, inspect the returned error rather than relying on a conversational answer.
Grok 4.3 pricing for redirected requests
xAI says requests sent to retired Grok 4 Fast IDs are billed at Grok 4.3 pricing: $1.25 per million input tokens and $2.50 per million output tokens. These are Grok 4.3 rates, not prices for a still-available Grok 4 Fast model. (docs.x.ai)
Bottom line: Don’t build a new integration around grok-4-fast-reasoning or grok-4-fast-non-reasoning. Use grok-4.3 explicitly with the xAI Responses API, and confirm success from the API response.