Grok 4.7 — reason, code and research
Grok 4.7 brings reasoning, image understanding and search to your apps. Try a prompt, inspect the answer and pay for the usage you need.
- Reasoning
- 500K context
- Image input
- Tools
- Streaming
Try Grok 4.7
Chat with Grok 4.7, attach images and continue the conversation. Adjust the parameters or inspect the full request and response as JSON.
grok-4.7$1.20 in · $3.60 out per 1M tokens
Leave an optional field blank to omit it. Use the JSON view to send null explicitly.
Estimated from current rates. Tool charges are additional.
Real Grok 4.7 output
A small code execution check with a returned result. This is an actual response, not a promised benchmark.
Use the code interpreter once to compute 13*17. Return the number.
221
Generated October 10, 2026 with low reasoning effort and one code interpreter call. Answers may vary.
Grok 4.7 pricing
Grok 4.7 token rates below are the current rates your account pays. Cached input has its own rate.
Examples use standard service, uncached input and no tools. At 200,000 input tokens, the long-context tier applies to the whole request; priority and fast service double token rates. Tools are charged separately.
| Category | Price · 1M tokens |
|---|---|
| Input | $1.20 |
| Output | $3.60 |
| Cached input | $0.30 |
What is Grok 4.7?
Grok 4.7 is xAI’s reasoning model for text, code and image understanding. Its 500K-token context gives you room for long documents, source files and conversation history in one request.
Use Grok 4.7 to answer a question, call a tool or return structured data. Choose a reasoning effort and check the result against your examples.
Model ID: grok-4.7 · Input: text and images · Output: text · Formats: Chat Completions, Responses and Messages.
Two ways to use Grok 4.7
Try Grok 4.7 from your application or with help from a coding agent.
Call it from your code
Keep the request format your app already uses. Grok 4.7 accepts Chat Completions, Responses and Messages with the same SeedRouter key.
- 1Create an API key
- 2Set the base URL to https://api.seedrouter.ai/v1
- 3Send grok-4.7 as the model
Give your agent a starting point
Copy the prompt below. It asks your agent to show the request and an estimated cost before making a paid call.
- 1Copy the prompt
- 2Paste it into your coding agent
- 3Review the request before approving
What you can do with Grok 4.7
Start with a task you can check, then expand the workflow.
Review code
Ask Grok 4.7 to explain a change, find a possible bug or propose a focused patch. Run your tests before accepting the result.
Read images
Give Grok 4.7 a screenshot or chart alongside your question. Ask for the specific detail you need, with enough context to judge the answer.
Return structured data
Request JSON or a JSON schema from Grok 4.7. Keep the structure small and validate the returned values in your application.
Research with tools
Use Grok 4.7 with web or X search when your question depends on current information. Inspect the sources returned with the answer.
Control reasoning effort
Choose low, medium, high or xhigh effort for Grok 4.7. Compare answer quality and reported usage on examples that matter to you.
Reuse prompt context
Repeated prompt prefixes may receive cached-input pricing. Check the actual cache count in your Grok 4.7 usage instead of assuming every repeat is cached.
Build with Grok 4.7
Document analysis
Send relevant sections together and ask Grok 4.7 for a comparison, checklist or summary you can check against the original.
Coding assistance
Give Grok 4.7 a failing example, the expected behavior and the surrounding code. Ask for a limited change with an explanation.
Tool-assisted answers
Let Grok 4.7 choose a function or run a calculation, then return the result to your app. Keep permissions in your application.
Start with Grok 4.7
- 01
Create your key
Sign in and create a key. Keep it on your server or in a local environment variable.
- 02
Choose your format
Try Grok 4.7 in the playground. Select Chat Completions, Responses or Messages to match your client.
- 03
Adjust the request
Add your context and parameters. Start with a small output limit while checking the response.
- 04
Check usage
Review the Grok 4.7 answer and usage. Your account records show the charge, including tools.
Grok 4.7 capabilities at a glance
Available formats and the current limits that affect your integration.
- Chat Completions
- Native request format
- Responses
- Native request format
- Messages
- Native request format
- Images
- Image inputs, text answers
- Streaming
- JSON and server-sent events
- Cached input
- Separate token rate
- Tools
- Functions, web search, X search and code execution
- Failed requests
- Not charged
Response ID continuation, stored-response operations, deferred Chat and file attachments are currently unavailable. Image generation and server-side tool discovery have not passed capability verification.
Grok 4.7 limits to plan for
Context
Grok 4.7 has a 500K-token context. Input, conversation history and generated output need room within that window.
Visible output
The default Grok 4.7 output limit is 128,000 visible tokens. Reasoning and function-call tokens are counted separately from that limit.
Long prompts
At 200,000 input tokens, Grok 4.7 uses the long-context price tier. Cached tokens still count toward this threshold.
Conversation history
For another Grok 4.7 turn, send the previous input and all returned output items, including encrypted reasoning where present.
Use Grok 4.7 on SeedRouter
One account
Call Grok 4.7 and your other models with the same account and API key.
Pay for usage
Use your balance as you go. Check tokens and tool charges in your request history.
Live token prices
The Grok 4.7 token rates displayed here are the current rates your account pays.
Failures are free
A Grok 4.7 request that returns an error is not charged.
Cache accounting
Cached input is part of total input in Chat and Responses. In Messages, cache-read input is reported separately.
Reasoning usage
Reasoning contributes to the Grok 4.7 bill. Use the format’s usage fields without adding the same reasoning tokens twice.
Streaming
Set stream to true to read Grok 4.7 output as it arrives. Keep the final usage event for accounting.
Related models
Other models available with the same key.
Call DeepSeek V4.1 Flash with the OpenAI or Anthropic SDK. 1M-token context, thinking on or off, image input, live peak and off-peak pricing and a playground.
View pricingCall Kimi K3 with the OpenAI or Anthropic SDK. 1M-token context, always-on reasoning with low, high or max effort, live per-token pricing and a playground.
View pricingCall Claude Fable 5 with the Anthropic or OpenAI SDK. Deep reasoning, 1M-token context, forced tool use, live per-token pricing and a playground to test it.
View pricingCall Claude Fable 5.1 with the Anthropic or OpenAI SDK. Deep reasoning, 1M-token context, cheap cached reads, live pricing and a playground to try it today.
View pricingTry Claude Haiku 5.5 for classification, extraction and summaries. See live token prices, test your prompts, and connect your app with the Messages API.
View pricingCall Claude Opus 5.5 with the Anthropic or OpenAI SDK. 1M-token context, adaptive thinking, live per-token pricing and a playground you can try right now.
View pricingGrok 4.7 guides
Read the pricing breakdown, including caching and reasoning usage.
Grok 4.7 FAQ
What is Grok 4.7?+
Grok 4.7 is an xAI model for reasoning, coding and understanding text and images. It supports a 500K-token context, structured output and tool calls, so you can use it for both direct answers and application workflows.
How much does Grok 4.7 cost?+
Standard token prices are $1.20 / 1M for input and $3.60 / 1M for output. Cached input and long-context requests have separate rates, shown above. Priority or fast service doubles token rates; tool use adds charges according to the work performed.
Can I change the reasoning effort?+
Yes. Grok 4.7 offers low, medium, high and xhigh, with high as the default. Chat uses reasoning_effort; Responses accepts a reasoning configuration. Higher effort can change the amount of reasoning performed, so check reported usage.
How does cached input affect the bill?+
When Grok 4.7 reports cached input, those tokens use the cached-input rate rather than the ordinary input rate. Cache hits are not guaranteed. The pricing guide shows how to calculate a request without counting cached tokens twice.
How do I continue a conversation?+
Send the full Grok 4.7 conversation history, including reasoning and tool output items. Response ID continuation and stored-response retrieval are unavailable; accepting store does not establish support for those operations.
Which request formats can I use?+
Grok 4.7 accepts Chat Completions, Responses and Messages. Use grok-4.7 at the matching endpoint. Each format has its own parameters, shown separately in the playground and API reference.
Will a failed Grok 4.7 request use my balance?+
A Grok 4.7 request that returns an error is not charged. Successful requests are billed for reported usage, including reasoning and billable tools. Review the charge in your usage records.
Try your task with Grok 4.7
Start with one prompt. Inspect the answer, compare usage and take the request into your code.