Grok Pricing 2026

Grok offers paid plans starting at $8/month (X Premium) and going up to $300/month (SuperGrok Heavy). Standalone xAI plans are SuperGrok Lite at $10/month, SuperGrok at $30/month, and SuperGrok Heavy at $300/month. X platform bundles are X Premium at $8/month and X Premium+ at $40/month.

The cheapest paid Grok access is X Premium at $8/month, which bundles basic Grok access with X platform features. For a standalone Grok subscription without needing X, SuperGrok Lite at $10/month is the entry option.

Grok does have a free tier available at grok.com with rate-limited access. Paid plans unlock higher usage limits, additional models, and features like DeepSearch and Big Brain mode.

SuperGrok costs $30/month and provides full Grok access with unlimited conversations, DeepSearch, and Big Brain mode. SuperGrok Heavy costs $300/month and adds access to the Grok Heavy model, multi-agent architecture, and elevated limits for demanding or complex workloads.

SuperGrok Lite is a $10/month standalone xAI plan launched in March 2026. It sits between the free tier and the full SuperGrok plan, offering a lower-cost entry point for users who want a dedicated Grok subscription without the full feature set of SuperGrok at $30/month.

X Premium ($8/month) includes basic Grok access bundled with X platform features. X Premium+ ($40/month) provides enhanced Grok access with higher usage limits within X, plus a fully ad-free X experience. Neither X bundle matches the unlimited usage of the standalone SuperGrok plan at $30/month.

Grok 4 is marketed at $3/million input tokens and $15/million output tokens, but the model uses reasoning (thinking) tokens during complex tasks that are billed on top of the base rate. For reasoning-heavy tasks, actual costs can be substantially higher than the headline price suggests.

Grok 3 mini is available at approximately $0.35/1M tokens on xAI's API, which is competitive with other smaller models. However, Grok 4 (the flagship) is described by some analysts as potentially the second most expensive model on the market when thinking token costs are factored in. Additionally, Grok's tendency to produce longer responses than competitors can increase effective per-task costs.

SuperGrok Heavy is the highest-tier standalone Grok subscription at $300/month, designed for users requiring maximum compute allocation and the heaviest AI workloads. It sits above the standard SuperGrok plan ($30/month) in the tier hierarchy.

Grok 4 is advertised at $3/million input tokens and $15/million output tokens — similar to Claude Sonnet 4. However, the model burns significant thinking tokens that substantially increase the effective cost. Community analysis suggests Grok 4 may be one of the most expensive models in practice despite its competitive headline pricing. Monitor actual token consumption carefully before running production workloads.

It depends on the model and use case. Grok 3 mini is available at $0.35/1M tokens — competitive with open-source alternatives. However, Grok 4's thinking token overhead can make it significantly more expensive than its headline rate suggests. Additionally, Grok models historically tend to produce longer responses than competing models, which increases per-task token costs even at equivalent per-token rates.

The SuperGrok Heavy plan at $300/month is positioned for users who need the highest-intensity Grok access. Based on available source data, this tier is designed for heavy power users, but specific feature differentiation from SuperGrok ($30/month) beyond usage volume is not well-documented in community sources. The 10x price premium over SuperGrok suggests it is intended for enterprise-level or extremely high-volume individual use.

Yes, and this significantly affects the true cost. Grok 4 advertises $3/million input and $15/million output tokens, but the model burns additional thinking tokens during its reasoning process that are billed on top of the headline rates. For complex tasks requiring deep reasoning, effective costs can be substantially higher than the listed prices — some analysts consider Grok 4 potentially the most expensive model on the market when thinking token consumption is included.

Not always. The advertised $2/M input and $6/M output for Grok 4.5 applies only up to 200k context; above that, pricing jumps to $6/$12 per million tokens. Thinking tokens on Grok 4 also inflate costs significantly beyond the headline rate. For agentic tasks where cached tokens represent 80–90% of usage, cache pricing adds substantially to real-world bills.

Yes. Grok's free tier provides access to the core model with basic queries and search. Key limits are ~10 searches and ~50 basic queries per 2-hour rolling window. The free tier excludes Big Brain Mode, Voice Mode, comprehensive DeepSearch, unlimited image generation, and Think Mode. Upgrade to SuperGrok ($30/month) if you regularly hit these limits.

Based on current published pricing, all Grok plans are listed at monthly rates only with no annual billing option or discount published. Verify on the xAI pricing page for the latest billing options.