xAI General reasoning & knowledge work
Grok 4.6
Use sustained planning when a simple completion will not hold the whole task.
Grok 4.6 is an xAI route for high-quality coding, agentic execution and knowledge work. TokenHot lists text, image and document inputs, a 500K context value, reasoning, tool use and structured output. Use it for sustained planning and multimodal analysis while keeping product-level search or browsing behavior separate from the model route.
- Text + image + document
- 500K context
- Structured tool use
A model route is not the same as a consumer product feature. Search, browsing, account context and external actions depend on the endpoint and tools your application supplies.
AI Chat
Send a prompt and see the response here.
grok-4.6Start a conversation with this model and ask follow-up questions.
USD /1M tokens
Pricing
| Group | Input | Output | Cache Read |
|---|---|---|---|
| default | $0.84 | $2.52 | $0.21 |
Public list prices are shown in USD. Final charges may vary by account group and usage tier.
Overview
What is Grok 4.6?
Grok 4.6 fits tasks that need several sources, careful planning or code and document context in one request. Name the sources and expected artifact so a broad answer does not blur facts with recommendations.
If a workflow needs real-time information, make retrieval an explicit step and store the returned source metadata. Do not describe a product-level feature as an automatic property of the API model.
Separate facts and advice
Require source names, dates and a recommendation section so a fluent memo stays accountable.
Keep the patch narrow
Provide the issue, repository context and tests. Ask for a diff plan and a verification record.
Use documents as evidence
Pair a screenshot or document with a concrete question and preserve the relevant page or region.
From brief to deliverable
Put Grok 4.6 to work
Strategy and operations teams
Draft a source-linked decision memo
Compare supplied documents and return agreed facts, disagreements, recommendation and unanswered questions.
- Tools you need
- Document retrieval and a human decision owner.
- Keep in mind
- The model cannot establish a source's currentness or authority without external evidence.
View a task brief
Use the supplied named documents, dates, decision criteria and a memo structure. to plan draft a source-linked decision memo. Return a traceable memo with a clearly bounded recommendation., state the evidence limits and list the checks required before accepting the result. Do not claim an output or interaction was verified without direct evidence.
Engineering teams
Review a proposed code change
Read the issue, diff and tests to identify regressions, missing cases and the smallest follow-up checks.
- Tools you need
- Source access and CI or local tests for confirmation.
- Keep in mind
- Review comments are not proof that a test or deployment passed.
View a task brief
Use the supplied issue text, patch, test output and repository conventions. to plan review a proposed code change. Return prioritized review findings and a verification checklist., state the evidence limits and list the checks required before accepting the result. Do not claim an output or interaction was verified without direct evidence.
Product and frontend teams
Inspect a document or screenshot
Use visual evidence to identify concrete layout or content issues, then tie each observation to a browser or source check.
- Tools you need
- Image or document input and a browser for interaction testing.
- Keep in mind
- A static view cannot establish hidden state, focus order or network behavior.
View a task brief
Use the supplied a screenshot or document, expected behavior and relevant code. to plan inspect a document or screenshot. Return an actionable issue list and reproduction checklist., state the evidence limits and list the checks required before accepting the result. Do not claim an output or interaction was verified without direct evidence.
Capabilities in context
Key features
| Feature | What you get | Why it matters |
|---|---|---|
| Context capacity | 500K tokens shown in the catalog | Use the window for connected project material, but keep retrieval and output boundaries deliberate. |
| Input modes | Text, image and document inputs are listed | Use visual or document evidence with page, region or source identifiers. |
| Reasoning | Reasoning is listed by the catalog | Measure accepted work and latency on your own tasks rather than assuming a product label predicts quality. |
| Tool use | Tool use and structured output are listed | Keep retrieval, authorization and execution in application-owned functions. |
| Product boundary | Model route versus product tools | Search, browsing and account features need explicit endpoint support and should not be implied by the model name. |
The page uses TokenHot catalog metadata. Product-level search or browsing behavior is intentionally not presented as an automatic API guarantee.
Audience & fit
Where Grok 4.6 fits
Teams producing high-stakes memos
Use source-linked structure and human review for research, strategy and operational decisions.
Engineering teams
Pair code context with explicit tests and use the route for sustained review or planning rather than unbounded edits.
Product and frontend teams
Use image and document input for first-pass review, then verify interaction behavior in a browser.
Cost context
Plan the cost of a useful result
Separate retrieval cost
Search, document parsing and storage can add material cost before the model call. Track them as part of the workflow.
Keep output scoped
A broad context can produce a broad answer. Define the artifact and maximum response size before sending.
Compare accepted outcomes
Record corrections, source coverage, latency and total cost rather than relying on a first-response impression.
Use the current pricing table for available rates and account groups. Model IDs and prices come from TokenHot’s catalog; provider pricing and subscriptions are separate.
API
Code examples
curl --location --request POST 'https://api.tokenhot.ai/v1/responses' \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data-raw '{
"model": "grok-4.6",
"input": [
{
"role": "user",
"content": "What is xAI?"
}
],
"tools": [
{
"type": "web_search"
}
]
}'FAQ
Frequently asked questions
What is Grok 4.6 best suited for?
Grok 4.6 is suited to the input, output, and capability types listed on this page. Test production workloads before choosing it for a critical system.
How is Grok 4.6 priced?
Pricing is shown in the pricing section above. Actual cost depends on usage volume, account group, and the request payload.
How can I call Grok 4.6?
Use the model ID shown above with one of the supported API protocols. Code examples are provided when usage data is available.
Does Grok 4.6 automatically browse the web?
No product feature should be inferred from the model name. Verify whether your selected endpoint supports retrieval and configure it explicitly.
Can it analyze images and documents?
The catalog lists text, image and document inputs. Verify file limits and preserve page or region references in your application.
Does tool use mean it executes actions?
Your application must expose, authorize and execute tools. The model can return arguments but cannot claim an external action without tool evidence.
What is the effective context limit?
The current catalog shows 500K. Endpoint output budgets, media encoding and request construction can reduce the usable amount.
Get started
Build with Grok 4.6
API