On the API, Grok 4.7 costs $2.00 per million input tokens, $0.50 per million cached input tokens and $6.00 per million output tokens; prompts of 200k tokens or more are billed at $4.00, $1.00 and $12.00. Batch requests get a 20% discount and Priority Processing costs twice the standard rate. A faster variant in Cursor and Grok Build costs twice the standard rate.Official source 1 ↗
Available as
Available in Cursor and Grok Build, through the Grok API, and via third-party coding harnesses, model routers and cloud platforms. · The API can be called with the xAI Python SDK, the OpenAI Python client pointed at api.x.ai, the Vercel AI SDK or plain HTTP.Official source 1 ↗Official source 1 ↗
Grok 4.7 is SpaceXSI's flagship language model for coding and knowledge work, released on 21 September 2026. It has a 500k-token context window, configurable reasoning effort and agentic tool calling, and is available in Cursor, Grok Build and the Grok API.
Built for longer coding tasks. Grok 4.7 uses a larger base model than Grok 4.6 and was trained on tasks that take many hours, with a focus on checking its own work and handling longer context.Official source 1 ↗
Adjustable reasoning effort. The reasoning_effort parameter accepts low, medium, high (the default) and xhigh, trading latency for deeper thinking.Official source 1 ↗
Documents and presentations. SpaceXSI reports better results than Grok 4.6 when creating documents and presentations for professional tasks.Official source 1 ↗
Server-side tools. Requests can use server-side tools such as web search, X search, code execution, collections search over your documents and remote MCP servers, each billed per call or by tokens.Official source 1 ↗
Encrypted reasoning across turns. On the Responses API, Grok 4.7 returns encrypted reasoning content with every response, which you can pass back so the model keeps its reasoning between turns.Official source 1 ↗
Safety for cybersecurity work. The model has a new safeguard stack; on SpaceXSI's HackerBench v0.3 it let 3.3% of risky dual-use prompts through.Official source 1 ↗
Pricing & free limits
Grok API (standard)
$2.00 input / $6.00 output per 1M tokensPay per token
Cached input is $0.50 per 1M tokens. Once a prompt reaches 200k tokens, all tokens in the request are billed at $4.00 input, $1.00 cached and $12.00 output per 1M.Official source 1 ↗
Batch API
20% off standard token ratesPay per token, asynchronous
Requests are processed in the background, most within 24 hours, and do not count towards rate limits.Official source 1 ↗
Priority Processing
2x standard token ratesPay per token
Higher scheduling priority for lower latency on Chat Completions and Responses requests; billed at the priority rate only when the response confirms the priority tier.Official source 1 ↗
US regional endpoint
$2.20 input / $6.60 output per 1M tokensPay per token
Inference runs in the United States at 1.1x the global rates; $0.55 cached input, and $4.40 / $1.10 / $13.20 above 200k prompt tokens.Official source 1 ↗
Grok 4.7 Fast (Cursor and Grok Build)
$4.00 input / $12.00 output per 1M tokensBilled through your Cursor or Grok Build plan
The same model on faster infrastructure; $1.00 cached input below 200k prompt tokens, and $6.00 / $1.50 / $18.00 above. Not available on the public API or in Grok Build's free tier.Official source 1 ↗
Billing details
API usage is paid from prepaid credits or, on request, by monthly invoice. Auto top-up can buy more credits when the balance falls below a threshold, with a minimum top-up of $5.Official source 1 ↗
Server-side tools are charged on top of tokens, for example $5 per 1,000 web search calls. Requests found to violate the usage guidelines are still charged, with a $0.05 fee for violations caught before generation.Official source 1 ↗
Free access & limits
Grok 4.7 can be tried for free in Grok Build, but the free tier does not include the faster Grok 4.7 Fast variant.Official source 1 ↗ · Official source 2 ↗
Professional knowledge work such as drafting documents and presentations for legal, clinical or financial tasks.Official source 1 ↗
No current events without search. Grok 4.7's knowledge stops at May 2026; current information requires enabling the web or X search tools.Official source 1 ↗
Reasoning cannot be turned off. Reasoning cannot be disabled, and presencePenalty, frequencyPenalty and stop return an error on reasoning models.Official source 1 ↗
No logprobs. logprobs and top_logprobs are ignored on grok-4.20 and newer models.Official source 1 ↗
Requests stop when credits run out. Without invoiced billing, API requests are rejected once prepaid credits are used up.Official source 1 ↗
Submit a tool
