Grok context window 2026: 500K tokens on Grok 4.7, 1M on Grok 4.3
Updated: 9 hours ago

Updated: November 16, 2026
The current flagship model, Grok 4.7, has a 500,000-token context window in the xAI API, while the older and cheaper Grok 4.3 and the Grok 4.20 models still offer 1,000,000 tokens. Grok 4.6 and Grok 4.5 also sit at 500,000 tokens, and the coding model grok-build-0.1 is listed at 256,000. Every model on xAI's pricing page switches to a higher long-context rate once a single prompt reaches 200,000 tokens, and that higher rate then applies to every token in the request.
For people who use Grok in the grok.com website or the mobile apps, xAI does not publish a per-plan context window. Its consumer pages describe usage allowances, file upload caps and data controls, but no token figure for chats. The table below lists the context windows and token prices that xAI's developer documentation showed on November 16, 2026, followed by cost examples, the rules that decide what fills the window, and the memory and history controls in the Grok apps.
........
Model (API ID) | Context window | Input / cached / output under 200K | Input / cached / output from 200K |
grok-4.7 | 500,000 | $2.00 / $0.50 / $6.00 | $4.00 / $1.00 / $12.00 |
grok-4.6 | 500,000 | $2.00 / $0.50 / $6.00 | $4.00 / $1.00 / $12.00 |
grok-4.5 | 500,000 | $2.00 / $0.30 / $6.00 | $4.00 / $0.60 / $12.00 |
grok-4.3 | 1,000,000 | $1.25 / $0.20 / $2.50 | $2.50 / $0.40 / $5.00 |
grok-4.20-0309 (reasoning, non-reasoning, multi-agent) | 1,000,000 | $1.25 / $0.20 / $2.50 | $2.50 / $0.40 / $5.00 |
grok-build-0.1 | 256,000 | $1.00 / $0.20 / $2.00 | $2.00 / $0.40 / $4.00 |
........
Prices are in USD per 1 million tokens, as listed on xAI's official API pricing page on November 16, 2026.
··········
GROK 4.7 CONTEXT WINDOW AND OUTPUT LIMIT.
xAI released grok-4.7 on September 21, 2026, according to its API release notes, with a 500,000-token context window and the same pricing tiers as Grok 4.6, which arrived on August 12, 2026. The Grok 4.7 developer guide lists the window as 500,000 tokens and states "No text output limit", so the practical ceiling on a reply is whatever room the prompt leaves inside the window. The model's knowledge cut-off is May 2026.
Grok 4.7 is a reasoning model with four effort settings: low, medium, high (the default) and xhigh. On the Responses API it always returns its reasoning in encrypted form, which lets an application pass that state back on the next turn without seeing the reasoning text. A faster variant, Grok 4.7 Fast, exists only inside Cursor and xAI's own coding agent Grok Build; it is not offered on the public API, and xAI says it costs twice the standard token rates, or 1.5 times for long-context requests.
The jump from Grok 4.3 to Grok 4.6 and 4.7 reduced the maximum window from 1 million to 500,000 tokens. Developers who need to send more than half a million tokens in one request still have two options in the current catalog: grok-4.3 and the grok-4.20-0309 family, all listed at 1,000,000 tokens and at lower per-token prices than Grok 4.7.
··········
GROK LONG-CONTEXT PRICING ABOVE 200,000 TOKENS.
The context window and the price are linked by one rule on xAI's pricing page: once a request's prompt reaches the long-context threshold of 200,000 tokens, the long-context rates apply to all tokens in that request, input and output alike. For Grok 4.7 this doubles input from $2.00 to $4.00, cached input from $0.50 to $1.00 and output from $6.00 to $12.00 per million tokens. Grok 4.3 follows the same pattern from a lower base, moving from $1.25 / $0.20 / $2.50 to $2.50 / $0.40 / $5.00.
Three other price modifiers affect long requests. The Batch API gives a 20% discount on input, output, cached and reasoning tokens for Grok 4.7, 4.6, 4.5, 4.3 and the Grok 4.20 0309 models. Priority Processing bills all token types at twice the standard rate when the response reports the priority service tier. The US regional endpoint, available for Grok 4.7 and Grok 4.6, adds a 10% premium, which puts Grok 4.7 at $2.20 / $0.55 / $6.60 below 200,000 prompt tokens and $4.40 / $1.10 / $13.20 above it.
The examples below were computed with these official rates. Each request produces 10,000 output tokens; the cached example assumes 250,000 of the 300,000 prompt tokens are served from cache.
........
Request | Model | Rate tier | Cost per request |
190,000 input + 10,000 output | grok-4.7 | Standard | $0.44 |
300,000 input + 10,000 output | grok-4.7 | Long context | $1.32 |
300,000 input (250,000 cached) + 10,000 output | grok-4.7 | Long context | $0.57 |
300,000 input + 10,000 output, Batch API | grok-4.7 | Long context, 20% off | $1.06 |
190,000 input + 10,000 output | grok-4.3 | Standard | $0.26 |
300,000 input + 10,000 output | grok-4.3 | Long context | $0.80 |
800,000 input + 10,000 output | grok-4.3 | Long context | $2.05 |
........
Crossing the threshold triples the cost of the Grok 4.7 example, from $0.44 to $1.32, although the prompt grows by about 58%. Keeping a long prompt just under 200,000 tokens, or reusing a cached prefix, has a larger effect on the bill than the choice between Grok 4.7 and Grok 4.6, which share the same prices.
··········
[[ADPLUS_300x250_1]]
WHAT COUNTS TOWARD THE GROK CONTEXT WINDOW.
The context window is a per-request budget. Everything the model receives in a call counts: the system prompt, the earlier turns of the conversation that the application sends again, attached file content, tool results and the model's own reply. With reasoning models, the reasoning tokens are billed as well, and xAI's Batch discount explicitly covers them as a separate token type. An application that keeps a long chat going therefore spends more of the window with every turn, because each earlier turn it sends again counts once more toward the limit.
Tools add tokens and, in most cases, a separate fee. Search results and fetched content enter the prompt and consume part of the window. On xAI's pricing page, Web Search costs $5 per 1,000 calls, X Search $5 per 1,000 posts and $10 per 1,000 profiles, Code Execution $5 per 1,000 calls, File Attachments search $5 per 1,000 calls and Collections Search $2.50 per 1,000 calls. Image understanding, X video understanding and remote MCP tools carry no call fee and are billed only through the tokens they add.
For long document work, retrieval can keep requests below the 200,000-token threshold. Collections Search retrieves the relevant passages from a stored document set at $2.50 per 1,000 calls, which is often cheaper than sending an entire archive in every prompt at the long-context rate.
··········
GROK CONTEXT, MEMORY AND CHAT HISTORY IN THE GROK APPS.
The Grok website and apps do not show a token counter, and xAI's consumer FAQ and plan pages publish no context window per plan. The plans currently listed on xAI's pricing page are Free, Plus at $8 a month, Premium at $30, Super at $100 and Ultra at $200, with usage described as multiples of the Plus allowance (4x for Premium, 15x for Super, 50x for Ultra) in a weekly pool shared between Grok and Cursor. Our guide to Grok plans and pricing covers what each plan includes.
File uploads have their own limits in the apps: xAI's FAQ lists up to 150 MB per file, up to about 100 files per upload on the web and up to 20 on Android, with limits that "can vary slightly by platform or subscription".
The data controls that shape what Grok carries between chats are documented in xAI's consumer FAQ. The "Personalize Grok using X" setting (Settings > Data Controls on mobile, Settings > Data on the web) decides whether your X data and interactions personalize answers. Private Chat, opened from the ghost icon in the top right corner, hides the conversation from your history, deletes it from xAI systems within 30 days and keeps it out of model training. Deleted conversations are also removed within 30 days, unless they are de-identified or must be kept for safety, security or legal reasons, and the "Improve the model" setting controls whether new conversations are used for training. None of these settings changes the number of tokens the model can read in one request.
··········
HOW TO CHOOSE A GROK MODEL FOR LONG INPUTS.
For prompts below 200,000 tokens, Grok 4.7 is xAI's current flagship and costs $2.00 input and $6.00 output per million tokens. For prompts between 200,000 and 500,000 tokens, Grok 4.7 still accepts the request, but every token moves to the long-context rate, so caching the stable part of the prompt and using the Batch API for non-urgent jobs make the largest difference. For prompts above 500,000 tokens, only grok-4.3 and the grok-4.20-0309 models accept the request, with windows of 1,000,000 tokens.
Developers can check token use in the usage data returned with each API response and compare it with the 200,000-token threshold before sending large jobs. The full list of Grok models, including image, video and voice models, is in our overview of all Grok models and when to use each, and the Grok 4.3 model guide covers the 1M-token model in more detail.
··········
GROK CONTEXT WINDOW LIMITS IN NOVEMBER 2026.
xAI's documentation sets three numbers that define Grok's context today: 500,000 tokens for Grok 4.7, 4.6 and 4.5, 1,000,000 tokens for Grok 4.3 and the Grok 4.20 0309 models, and a 200,000-token threshold above which every token in the request costs twice as much. Grok 4.7 has no separate cap on text output, so long replies compete with the prompt for the same 500,000 tokens.
The consumer side is less documented. xAI publishes plan prices, usage multiples, file limits and data controls for the Grok apps, but no context window per plan, and no official page confirms which model length applies inside a chat. Readers who need a guaranteed token budget can rely on the API figures above, which xAI documents model by model.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org

