Grok 4.7 model IDs by channel: xAI API, OpenRouter, Vercel AI Gateway, Cloudflare, Cursor and Vertex

What is the Grok 4.7 model ID on each channel?

Short answer: grok-4.7 on the xAI API, x-ai/grok-4.7 on OpenRouter, spacexai/grok-4.7 on Vercel AI Gateway and xai/grok-4.7 on Cloudflare. Set the ID per channel, because the strings don't carry over.

Table A: Grok 4.7 IDs and how you call them (last checked 8 Oct 2026)

Quick answers:

• What is Grok 4.7's model ID on OpenRouter? x-ai/grok-4.7, or the alias ~x-ai/grok-latest. OpenRouter's canonical slug for it is x-ai/grok-4.7-20260916.
• What is Grok 4.7's model ID on Vercel AI Gateway? spacexai/grok-4.7. Vercel lists SpaceXAI as its only provider.
• What is Grok 4.7's model ID on Cloudflare? xai/grok-4.7, called with env.AI.run('xai/grok-4.7', { messages }) in a Worker or through Cloudflare's Chat Completions REST endpoint.

Which prefix belongs to which platform?

Short answer: x-ai/ is OpenRouter, spacexai/ is Vercel AI Gateway and xai/ is Cloudflare. The xAI API itself takes the bare grok-4.7.

xai/grok-4.7 isn't unique to Cloudflare: LiteLLM uses the same string, and xAI's Vertex guide uses it as an example of a publisher-prefixed name. Don't assume a prefixed ID from one platform works on another.

What does Grok 4.7 cost on each channel?

Short answer: xAI's list price is $2 input and $6 output per million tokens below 200K prompt tokens. On 8 Oct 2026, OpenRouter, Cloudflare and Cursor's standard rate all matched xAI's $2 and $6.

Table B: Grok 4.7 prices per 1M tokens (last checked 8 Oct 2026)

Check Vercel's own model page for its current Grok 4.7 price.

More price details:

• OpenRouter priority endpoints (xai/priority, xai/zdr/priority) listed $4.00 input and $12.00 output, or $8.00 and $24.00 above 200K, on 8 Oct 2026.
• Web search through OpenRouter is listed at $0.005 per call, the same $5 per 1,000 calls as xAI's own web_search price.
• xAI's US regional endpoint adds 10% to every token price.

How much is Grok 4.7 on OpenRouter? On 8 Oct 2026, OpenRouter listed x-ai/grok-4.7 at $2.00 per million input tokens and $6.00 per million output tokens, with cache reads at $0.50, the same as xAI's list price. Check OpenRouter's model page for today's price.

Why is the long-context threshold 200K on xAI but 256K in Cursor?

Short answer: Each platform sets its own rule. xAI bills every token at the higher rate once the prompt reaches 200K tokens; Cursor switches to long-context billing only when input exceeds 256K tokens, and counts input length only.

Every channel lists the same 500,000-token context window. The difference is only where the higher price starts.

Why does Cursor charge long-context rates later than xAI? Cursor's Grok 4.7 page uses a 256K standard window and bills long context only when input exceeds 256K tokens. xAI's API uses a 200K threshold and applies the higher rate to every token in the request.

Is Grok 4.7 on Google Vertex AI?

Short answer: Yes. Google's partner-model docs, last updated 7 Oct 2026, list Grok 4.7 and say to use the model name grok-4.7. xAI's Vertex guide shows xai/grok-4.7 only as an example of a publisher-prefixed name, so use the name Model Garden shows.

What each source says:

On Vertex, use the model name shown on Google's partner-model page and in Model Garden.

Which platforms does xAI name for Grok 4.7? xAI's Grok 4.7 model card names the SpaceXAI API, Grok Build, Cursor (all plans), the Office add-ins for Word, PowerPoint and Excel, and the gateways "OpenRouter, Vercel, Cloudflare, Snowflake, Databricks Mosaic, and others". It doesn't name Amazon Bedrock, Google Vertex AI or Microsoft Foundry. Amazon Bedrock and Google's Vertex docs now list Grok 4.7 themselves (Bedrock model ID xai.grok-4.7, launch date 28 Sep 2026).

Why does Vercel still list grok-4.1-fast models?

Short answer: Because Google Vertex AI serves them, not xAI. On Vercel's model pages, the only provider for spacexai/grok-4.1-fast-reasoning and spacexai/grok-4.1-fast-non-reasoning is Google Vertex AI, so xAI's own retirement redirect doesn't apply.

xAI retired its hyphenated slugs grok-4-1-fast-reasoning and grok-4-1-fast-non-reasoning from its own API on 15 May 2026, and requests to them now go to grok-4.3. Google's partner-model docs still list grok-4.1-fast-reasoning and grok-4.1-fast-non-reasoning, now marked Deprecated. Vercel lists them at $0.20 input and $0.50 output per million tokens, with a 1M context window.

So the same family name means two different things: on xAI's API it's a redirect, and on Vercel it's a model served through Google. For xAI's full retirement table, see what happened to retired Grok models.

Do request parameters behave the same on every channel?

Short answer: Not according to the catalogs. OpenRouter lists logprobs and top_logprobs as supported for x-ai/grok-4.7, and Vercel lists stop, but xAI's docs say logprobs is silently ignored and stop returns an error on reasoning models.

The gateway docs don't say what happens when you send these parameters, so the safe choice is to follow xAI's rules everywhere. The full list of Grok parameter quirks is in Grok function calling and structured outputs.

Where can I use Grok 4.7 Fast?

Short answer: Only in Cursor and Grok Build. xAI says Grok 4.7 Fast isn't available on the public xAI API. It costs 2x the standard token rates, or 1.5x for long-context requests, according to xAI.

In Cursor, Fast is the default speed on Pro and higher plans, at $4 input, $1 cached input and $12 output per million tokens. Cursor's Start plan, which is available in India only, is fixed at medium effort and standard speed.

Which channel should I use for Grok 4.7?

Short answer: Use the platform you already have an account and billing with. If you need xAI features such as encrypted reasoning or the server-side tools exactly as xAI documents them, call the xAI API directly.

A few things to weigh:

• Price: gateway prices can differ from xAI's list price and change without notice. Note the date of any price you rely on.
• Long context: xAI's 200K rule and Cursor's 256K rule give different bills for the same prompt.
• Parameters: catalogs don't always match xAI's docs (see the table above).

To get started on the direct API, see the Grok API quickstart. For the same format applied to Jev, see Jev model IDs by channel.

FAQ

What is Grok 4.7's model ID on OpenRouter?

x-ai/grok-4.7, or the alias ~x-ai/grok-latest. OpenRouter's canonical slug for it is x-ai/grok-4.7-20260916.

What is Grok 4.7's model ID on Vercel AI Gateway?

spacexai/grok-4.7. Vercel lists SpaceXAI as its only provider.

What is Grok 4.7's model ID on Cloudflare?

xai/grok-4.7, called with env.AI.run('xai/grok-4.7', { messages }) in a Worker or through Cloudflare's Chat Completions REST endpoint.

How much is Grok 4.7 on OpenRouter?

On 8 Oct 2026, OpenRouter listed x-ai/grok-4.7 at $2.00 per million input tokens and $6.00 per million output tokens, with cache reads at $0.50, the same as xAI's list price. Check OpenRouter's model page for today's price.

Why does Cursor charge long-context rates later than xAI?

Cursor's Grok 4.7 page uses a 256K standard window and bills long context only when input exceeds 256K tokens. xAI's API uses a 200K threshold and applies the higher rate to every token in the request.

Is Grok 4.7 on Google Vertex AI?

Yes. Google's partner-model docs, last updated 7 Oct 2026, list Grok 4.7 and say to use the model name grok-4.7. xAI's Vertex guide shows xai/grok-4.7 only as an example of a publisher-prefixed name, so use the name Model Garden shows.