Grok Context Window 2026: Every Tier's Token Limit Explained

Last Updated: August 1, 2026

Grok's newest flagship model has its smallest context window: Grok 4.5, launched July 8, 2026, carries a 500K token context window at the API level, while older models Grok 4.20 and Grok 4.1 Fast offer 2M tokens, and Grok 4.3 offers 1M token context per verified xAI documentation reviewed July 10, 2026. That counterintuitive pattern - newer model, smaller window - is the most important thing to understand about Grok context windows in 2026 before making any subscription or API decision.

The second critical distinction: API context windows are the technical maximum. The consumer web interface at grok.com and x.com/i/grok enforces a smaller working context in practice. Free tier users work with approximately 128K tokens regardless of what the API specs say. SuperGrok subscribers get larger context, but the exact window depends on which model their plan is currently serving during Grok 4.5's staged rollout.

This guide covers every Grok context window figure for August 2026 - by model, by subscription tier, and by API access - with every specification linked to a named primary source.

🎯 Before you read on - we put together a free 2026 AI Tools Cheat Sheet covering the tools business leaders are actually using right now. Get it instantly when you subscribe to AI Business Weekly.

Get the Free Cheat Sheet

Sources: xAI official API documentation, AI Toolbox Grok models explained July 2026, FelloAI Grok pricing July 2026, Rapidevelopers Grok 4 API limits July 2026

A context window is the total amount of text an AI model can read and consider at once - your prompt, any documents you upload, the conversation history, and the model's response all combined, measured in tokens where one token equals roughly three-quarters of a word in English.

A 500K token context window can hold approximately 375,000 words - roughly three full-length novels or a large codebase. A 2M token window holds around 1.5 million words - enough to process entire research archives or massive codebases in a single session.

Why it matters practically:

Document analysis: If you are pasting a 200-page PDF into Grok for analysis, you need at least 150K-200K tokens of available context just for the document. With a 128K free-tier window, that document does not fit.

Long conversations: Every message in a conversation consumes context. A multi-hour research session with dozens of exchanges can exhaust a 128K window, causing the model to lose earlier context and produce inconsistent or forgetful responses.

Code review: A large codebase review requiring Grok to hold thousands of lines of code across multiple files simultaneously demands 500K or more in practice.

The consumer vs API distinction:

The context windows advertised for Grok models reflect API-level maximums. The consumer interface on grok.com enforces a smaller working context per AI Toolbox's July 2026 Grok models breakdown. For maximum context access, the xAI API is the appropriate path. Consumer subscriptions offer higher context than the free tier but do not deliver full API-spec context in every session.

For a plain-language explanation of context windows across all major AI platforms including ChatGPT, Claude, and Gemini, our what is a context window guide covers everything you need to know.

Grok 4.5's context window is 500K tokens at the API level - the smallest of any current Grok model and exactly half the 1M token window of its predecessor Grok 4.3 per xAI documentation verified July 10, 2026. The trade-off: Grok 4.5 is faster, cheaper per token on short to medium requests, and xAI claims approximately 4x better token efficiency versus Claude Opus 4.8, meaning it accomplishes more within a smaller window.

Grok 4.5 context window key facts:

Source: AIWiz UK Grok 4.5 review July 26, 2026, AI Reiter Grok 4.5 launch confirmation, Rapidevelopers xAI API limits July 2026

The counterintuitive context window story:

Most AI model upgrades increase context window alongside capability. Grok 4.5 went in the opposite direction - smaller context, higher capability within that context. xAI's reasoning: modern agentic coding workflows benefit more from speed and token efficiency than from a large context window. Grok 4.5 is designed as an agentic coding model, not a long-document analysis model. If your use case requires processing very long documents or maintaining extensive conversation history, Grok 4.3 at 1M tokens or Grok 4.1 Fast at 2M tokens may serve you better - and at lower per-token cost.

For our complete guide to Grok 4.5 pricing, benchmarks, and features, our Grok AI pricing guide covers the full picture updated for August 2026.

Across the current Grok model lineup, context windows range from 256K tokens on Grok 4 Heavy to 2M tokens on Grok 4.20 and Grok 4.1 Fast, with the newest flagship Grok 4.5 sitting at 500K per AI Toolbox's July 2026 Grok models breakdown.

Complete Grok model context window comparison:

Source: AI Toolbox Grok models explained, FelloAI Grok pricing July 2026

The model selection guide by context need:

Under 200K tokens (most professional workflows): Grok 4.5 is the optimal choice - fastest, most capable on reasoning and coding tasks, and cheapest below the 200K threshold where standard rates apply.

200K to 500K tokens (large documents, extended sessions): Grok 4.5 still works but rates double above 200K tokens. Grok 4.3 at 1M context and $1.25/$2.50 pricing becomes more cost-effective for requests regularly in this range.

500K to 2M tokens (very long documents, large codebases, research archives): Grok 4.1 Fast at $0.20/$0.50 with a 2M context window is the clear choice. It is the cheapest Grok API option with the largest available context window. For complex multi-step tasks in this range, Grok 4.20 adds agentic variants.

Maximum reasoning on bounded tasks: Grok 4 Heavy at 256K tokens through SuperGrok Heavy delivers the highest reasoning capability but smallest window in the lineup. It is designed for problems where reasoning depth matters more than document volume.

Important note on Grok 4 (deprecated):

Grok 4 was retired May 15, 2026. The grok-4 API slug is deprecated and now redirects to Grok 4.3 billing per Rapidevelopers' July 2026 API limits verification. Any code or applications using the grok-4 model string should be updated to grok-4.3 or grok-4.5 explicitly.

Grok subscription tiers access context windows through the models they serve, not through a separate context window setting, and the staged Grok 4.5 rollout means SuperGrok subscribers may be on different models at different times, making the effective context window variable until the rollout completes.

Consumer tier context windows - August 2026:

Free tier (~128K tokens):
The free tier on grok.com and x.com/i/grok provides approximately 128K tokens of working context with limited Grok 4.3 access, approximately 10 prompts per two-hour window, and no Grok 4.5 access. Sufficient for single-document analysis of standard business reports and typical conversational sessions. Insufficient for large document processing or extended research sessions.

SuperGrok Lite ($10/month):
Entry paid tier with higher limits than free but no Expert mode, no Grok Imagine, and limited context compared to full SuperGrok. Appropriate for light users who hit the free tier limits but do not need maximum context.

X Premium ($8/month) and X Premium+ ($40/month):
Bundled Grok access with X social features. X Premium provides more Grok access than free. X Premium+ receives Grok 4.5 in a staged rollout alongside SuperGrok. Neither provides confirmed full Grok 4.5 context today. X Premium+ costs $10 more per month than SuperGrok while providing less AI capability per FelloAI's July 2026 analysis.

SuperGrok ($30/month):
The primary professional tier. Effective context window is variable during the Grok 4.5 staged rollout. Via Grok 4.1 Fast access, users may get 2M token context. As Grok 4.5 rolls out, that shifts to 500K. DeepSearch, Big Brain mode, Expert mode, and Grok Imagine are included. For most professional workflows under 200K tokens, SuperGrok on Grok 4.5 delivers excellent results. For document-heavy workflows over 500K tokens, the API provides better context options.

SuperGrok Heavy ($300/month):
The only consumer plan with confirmed full Grok 4.5 access as of August 1, 2026. Provides both Grok 4.5 (500K context) and Grok 4 Heavy (256K context, 16-agent parallel reasoning). Current promotional rate of $99/month for the first three months makes evaluation more accessible. Maximum rate limits across all features.

Source: FelloAI Grok pricing July 2026, AI Toolbox Grok pricing 2026

For the complete Grok subscription tier breakdown including all pricing and feature comparisons, our Grok AI pricing guide covers every plan updated for August 2026.

The xAI API provides access to every current Grok model at their full context window specifications, billed per token rather than through a subscription, with Grok 4.5 at 500K tokens as the flagship and Grok 4.1 Fast at 2M tokens as the largest-context and most cost-efficient option per xAI official documentation.

xAI API context windows and pricing - August 2026:

Source: xAI official API documentation, AI Toolbox Grok pricing 2026, Rapidevelopers xAI API limits July 2026

API access notes for August 2026:

The grok-4 model slug is deprecated as of May 15, 2026. Any applications using grok-4 should update to grok-4.3 or grok-4.5 explicitly. The old $150/month free API credit program ended May 2025. xAI offers up to $175/month in free credits through its data-sharing program - enable in your xAI console settings to access. API billing is independent of consumer subscriptions. A SuperGrok subscription does not include API credits.

Tool costs on top of token pricing:

Per Rapidevelopers' July 2026 API limits document:

In my experience working with executives evaluating AI tools, the tool costs are the most consistently underestimated component of Grok API budgets. A single DeepSearch-enabled research query can trigger multiple web search tool calls, adding $0.005 to $0.025 per query on top of token costs. At volume, those tool costs become significant.

Grok 4.5 bills at double the standard rates for both input and output above a 200K token input threshold per AIWiz UK's July 26, 2026 review - meaning a request using 300K input tokens costs significantly more than three times a 100K token request.

The rate doubling calculation:

Standard Grok 4.5 rates: $2.00/M input, $6.00/M output.
Above 200K input tokens: $4.00/M input, $12.00/M output on the full request.

Example: A 300K token document analysis with a 5K token output:

The same request costs exactly double due to the threshold trigger. For organizations regularly processing documents over 200K tokens, this doubles Grok 4.5 API costs relative to shorter-context use cases.

The practical implication:

For document processing workflows regularly exceeding 200K tokens, Grok 4.3 at $1.25/$2.50 with a 1M context window and no published threshold doubling is more cost-effective than Grok 4.5 despite Grok 4.5 being the newer model. The rate threshold is the clearest example of why model selection should be based on use case rather than recency.

Across the current AI assistant landscape, Grok 4.1 Fast's 2M token context window leads all major commercial AI models, while Grok 4.5's 500K window is smaller than Claude's 200K-in-practice limit but above ChatGPT Plus's effective consumer context.

Context window comparison across major AI models - August 2026:

Source: AI Toolbox Grok models comparison, what is a context window guide

The honest competitive picture:

On raw context window size, Grok 4.1 Fast at 2M tokens matches Gemini 3.1 Pro and exceeds ChatGPT and Claude. But context window size is not the only variable in long-document performance. How accurately a model retrieves information from deep within a large context window, known as "needle in a haystack" performance, varies significantly between models regardless of maximum window size.

Grok 4.5's 500K context window positions it below Grok 4.1 Fast and Grok 4.3, roughly comparable to ChatGPT Plus, and above Claude's 200K consumer context. For the specific use case of agentic coding where Grok 4.5 is designed to excel, 500K tokens is sufficient for the majority of real-world codebases.

For a complete head-to-head comparison of Grok against ChatGPT across all capabilities, our Grok vs ChatGPT guide covers every feature and use case. For how Grok compares to Claude specifically, our Grok vs Claude guide covers the full picture.

For document-heavy workflows, the right Grok plan depends entirely on document size: under 200K tokens use SuperGrok on Grok 4.5, between 200K and 1M tokens use the xAI API with Grok 4.3, and above 1M tokens use the xAI API with Grok 4.1 Fast at $0.20/$0.50 per million tokens.

Decision framework by document size:

Single standard business document (10-50 pages, under 50K tokens):
SuperGrok at $30/month. Grok 4.5 handles this comfortably within standard rates. No need for API access or heavy-tier subscriptions.

Large report or multiple documents (50-200 pages, 50K-200K tokens):
SuperGrok at $30/month via Grok 4.5, still within standard rate territory. Or Grok 4.3 API at $1.25/$2.50 if you want confirmed context access without staged rollout uncertainty.

Very large documents or full document sets (200K-1M tokens):
xAI API with Grok 4.3 ($1.25/$2.50 per M tokens, 1M context) is the most cost-effective choice. Grok 4.5 above 200K triggers rate doubling, making it twice as expensive for equivalent context.

Entire research archives or massive codebases (1M-2M tokens):
xAI API with Grok 4.1 Fast ($0.20/$0.50 per M tokens, 2M context) is the clear choice. Cheapest Grok API option, largest context window. For agentic multi-step tasks in this range, Grok 4.20 adds reasoning and multi-agent variants at the same $1.25/$2.50 pricing.

Maximum reasoning on complex bounded tasks (under 256K tokens):
SuperGrok Heavy at $300/month ($99 promo for first three months) provides Grok 4 Heavy's 16-agent parallel reasoning. Best for problems where reasoning depth matters more than document volume.

Grok AI Pricing 2026: SuperGrok, Heavy and Free Plan
Every Grok subscription tier and API price updated for August 2026 with the SuperGrok Lite and Grok 4.5 changes.

What Is a Context Window? Plain Language Guide
Context windows explained without jargon - what they are, how to measure them, and what they mean for your workflow.

Grok vs ChatGPT: Full Comparison 2026
Head-to-head across context window, reasoning, pricing, and real-world use cases.

Grok vs Claude: Full Comparison 2026
How Grok's context window and capabilities compare to Claude's 200K window and enterprise strengths.

SuperGrok vs ChatGPT Plus: Is the $10 Gap Worth It?
The direct subscription comparison covering context window, features, and value.

Grok AI Statistics 2026
The complete Grok platform data including user counts, revenue, and market position.

AI Chatbots Comparison Guide 2026
Context windows and capabilities compared across ChatGPT, Claude, Gemini, Perplexity, and Grok.

AI Statistics 2026: The Complete Data Guide
The master hub for all AI market data including xAI and Grok platform statistics.

What is Grok's context window in 2026?
Grok's context window varies by model and access method. Grok 4.5, the current flagship launched July 8, 2026, has a 500K token context window at the API level. Grok 4.3 has a 1M token context window. Grok 4.20 and Grok 4.1 Fast both offer 2M token context windows. Grok 4 Heavy, exclusive to SuperGrok Heavy, has a 256K token context window. The free consumer tier on grok.com works with approximately 128K tokens in practice. API context windows are the technical maximum; the consumer web interface enforces smaller working contexts. Source: xAI documentation, AI Toolbox July 2026

What context window does Grok 4.5 have?
Grok 4.5 has a 500K token context window verified against xAI's docs on July 10, 2026 per Rapidevelopers. 500K tokens is approximately 375,000 words or roughly 1,500 pages of standard text. This is smaller than Grok 4.3's 1M context and significantly smaller than Grok 4.1 Fast's 2M context. The trade-off is speed and token efficiency. xAI designed Grok 4.5 as an agentic coding model where 500K tokens is sufficient for most real-world codebases and speed matters more than context volume. An important caveat: Grok 4.5 doubles its standard rates above 200K input tokens per AIWiz UK's July 26, 2026 verification, making large-context requests significantly more expensive. Source: xAI docs, AIWiz UK July 2026

Which Grok model has the largest context window?
Grok 4.20 and Grok 4.1 Fast both offer the largest context window in the Grok lineup at 2M tokens each per AI Toolbox's July 2026 Grok models breakdown. 2M tokens is approximately 1.5 million words, enough to process entire books, large codebases, or extensive research archives in a single session. Grok 4.1 Fast is the most cost-efficient option for large-context use cases at $0.20/$0.50 per million tokens - the lowest price of any current Grok API model. Both are available through the xAI API. Consumer subscriptions on SuperGrok may access 2M context via Grok 4.1 Fast depending on which model is served during the Grok 4.5 staged rollout. Source: AI Toolbox

How does Grok's context window compare to Claude and ChatGPT?
Claude Opus 4.8 and Claude Pro offer a 200K token context window. ChatGPT Plus on GPT-5.6 Sol offers a 1M token context window. Gemini Advanced on Gemini 3.1 Pro offers a 2M token context window. Grok's context window ranges from 256K (Grok 4 Heavy) to 500K (Grok 4.5) to 1M (Grok 4.3) to 2M (Grok 4.1 Fast and Grok 4.20) depending on which model you access. On raw maximum context, Grok 4.1 Fast at 2M tokens matches Gemini Advanced and exceeds Claude and ChatGPT Plus. The practical consumer context depends on which model your subscription serves at any given time during the Grok 4.5 staged rollout. Source: AI Toolbox Grok comparison

Does the free Grok tier have a context window limit?
Yes. The free tier on grok.com and x.com/i/grok works with approximately 128K tokens in practice, even though some underlying Grok models support larger windows at the API level. The free tier also limits users to approximately 10 prompts per two-hour window and does not provide access to Grok 4.5, DeepSearch, Expert mode, or Big Brain mode. For regular professional use involving documents over 80-100 pages or extended research sessions, the free tier creates meaningful friction through both its context limit and its prompt rate limits. SuperGrok Lite at $10/month is the lowest-cost option that removes the most restrictive free-tier limits. Source: AI Toolbox Grok pricing 2026

What happens when you exceed Grok 4.5's 200K token threshold?
Both the input and output rates for Grok 4.5 double for the entire request when the input exceeds 200K tokens per AIWiz UK's July 26, 2026 verification and Rapidevelopers' July 10, 2026 API limits document. Standard rates are $2.00/$6.00 per million tokens. Above 200K input tokens the rates become approximately $4.00/$12.00 per million tokens. A 300K token document analysis that would cost $0.63 at standard rates costs approximately $1.26 at the above-threshold rate. For workflows regularly processing documents above 200K tokens, Grok 4.3 at $1.25/$2.50 with a 1M context window is more cost-effective than Grok 4.5 despite being an older model. Source: AIWiz UK, Rapidevelopers

Which Grok plan is best for processing large documents?
It depends on document size. For documents under 200K tokens, SuperGrok at $30/month via Grok 4.5 is sufficient for most professional workflows. For documents between 200K and 1M tokens, the xAI API with Grok 4.3 at $1.25/$2.50 per million tokens and 1M context is more cost-effective than Grok 4.5's doubled above-threshold rates. For documents or archives between 1M and 2M tokens, the xAI API with Grok 4.1 Fast at $0.20/$0.50 per million tokens with 2M context is the clear choice - cheapest Grok model with the largest available context. For complex multi-step reasoning on bounded tasks under 256K tokens where reasoning depth matters more than document volume, SuperGrok Heavy at $300/month ($99/month promotional rate for first three months) adds Grok 4 Heavy's 16-agent parallel reasoning.

Is Grok 4 still available in 2026?
No. Grok 4 was retired on May 15, 2026. The grok-4 API model slug is deprecated and now redirects to Grok 4.3 billing per Rapidevelopers' July 10, 2026 API limits verification. Any code or applications using the grok-4 model string will be automatically redirected to Grok 4.3 - update explicitly to grok-4.3 or grok-4.5 for predictable billing and behavior. Grok 4.3 with its 1M context window and Grok 4.5 with its 500K context window are the primary current models replacing Grok 4 in most workflows. Source: Rapidevelopers July 2026

Grok's context window story in August 2026 has one counterintuitive finding that changes how most users think about their model selection.

The newest and most capable Grok model - Grok 4.5 - has the smallest context window in the lineup at 500K tokens. An older, cheaper model - Grok 4.1 Fast - has the largest context window at 2M tokens and costs ten times less per million tokens. The model hierarchy by capability and the model hierarchy by context window point in opposite directions.

This is not a flaw in xAI's product strategy. It reflects a deliberate design choice: Grok 4.5 was built for agentic coding and speed-sensitive tasks where 500K tokens is sufficient and token efficiency matters more than maximum context. Grok 4.1 Fast was built for high-volume, large-context API use cases where price-per-token and context volume are the primary concerns.

The practical guidance is straightforward. Use Grok 4.5 when you need the most capable model on coding, reasoning, and agentic tasks and your inputs stay under 200K tokens. Use Grok 4.3 when you need 1M context at a reasonable price without Grok 4.5's rate-doubling threshold. Use Grok 4.1 Fast when you need the largest context at the lowest cost. Use Grok 4 Heavy through SuperGrok Heavy when you need the deepest multi-agent reasoning on bounded problems.

The subscription versus API decision matters too. Consumer subscriptions give you easy access without per-token billing. The API gives you access to every model at full context specifications with predictable per-token cost. For document-heavy professional workflows, the API with the right model for your context size is almost always the better economic choice.