Grok API and Subscription Pricing Guide
“Grok pricing” can mean metered xAI API usage or access through a consumer subscription. This page separates those products and lists the currently tracked API variants with a reproducible workload estimate.
Current API price snapshot
Up to 12 tracked language models are shown. Channels are selected by lowest input rate normalized to USD, not lowest total cost. Examples use 1M input plus 250K output tokens, without additional cache, batch or context-tier adjustments or tool fees. Processing-mode variants are not interchangeable; no total is shown when output pricing is unknown.
| Model | Lowest-input-rate channel | Input / 1M | Output / 1M | Example cost (USD) | Channels | Selected price verified |
|---|---|---|---|---|---|---|
| Grok 4.7 | XycAi | ¥4.00 | ¥12.01 | $1.015 | 3 | Oct 17, 2026 |
| Grok 4.6 | XycAi | ¥4.00 | ¥12.01 | $1.015 | 3 | Oct 17, 2026 |
| Grok 4.5 | XycAi | ¥4.00 | ¥12.01 | $1.015 | 3 | Oct 17, 2026 |
| Grok | OpenRouter | $2.00 | $6.00 | $3.50 | 1 | Oct 17, 2026 |
| Grok Build 0.1 | OpenRouter | $1.00 | $2.00 | $1.50 | 1 | Oct 17, 2026 |
| Grok 4.3 Batch | OpenRouter | $1.00 | $2.00 | $1.50 | 1 | Oct 17, 2026 |
| Grok 4.3 | XycAi | ¥2.50 | ¥5.00 | $0.5438 | 3 | Oct 17, 2026 |
| Grok 4.20 Multi Agent | Grok / X.AI | $1.25 | $2.50 | $1.875 | 2 | Oct 17, 2026 |
| Grok 4.20 | OpenRouter | $1.25 | $2.50 | $1.875 | 1 | Oct 17, 2026 |
| Grok 4.20 Non Reasoning | Grok / X.AI | $1.25 | $2.50 | $1.875 | 1 | Oct 17, 2026 |
| Grok 4.20 Reasoning | Grok / X.AI | $1.25 | $2.50 | $1.875 | 1 | Oct 17, 2026 |
| Grok Code Fast 1 | Grok / X.AI | $1.00 | $2.00 | $1.50 | 1 | Oct 17, 2026 |
Choose the right billing product
Use API pricing for applications, agents, and batch workloads. Use subscription pricing for the end-user Grok experience. A subscription should not be treated as interchangeable API credit.
Variant names matter
Fast, mini, reasoning, and dated Grok variants can have different token rates and limits. Match the exact model identifier used by your integration before comparing totals.
Cost formula: include reasoning and search
Grok API cost = ordinary input ÷ 1M × I + cached input ÷ 1M × C + billable output ÷ 1M × O + separately billed search and tool usage. Reasoning tokens are billed as output: if the output total already includes reasoning, do not add it twice. Live search and server-side tools have their own billing units, which may be calls or retrieved items; use the current tool entry. Consumer subscription access is a separate entitlement, not an interchangeable API credit balance; check any subscription-linked API allowance explicitly.
Usage example: a reasoning agent with search
Assume 1,000 agent requests with measured totals of 2M input tokens, including 0.5M cache hits, plus 0.25M visible completion and 0.75M reasoning tokens. Assume all search-result tokens are already in those input totals, no batch discount, and 200 billable web-search calls on a tool priced per call. The illustration costs 1.5 × I + 0.5 × C + 1 × O + 200 × S, where S is that tool’s current per-call rate. If the selected search tool bills retrieved items instead, replace the last term with the measured item count and its rate.
Easy-to-miss costs: hidden output and repeated tools
A short final answer can still carry substantial billed reasoning output. One user request can trigger multiple search or tool operations, and tool results enlarge later prompts. Cache-hit rates apply only to confirmed hits, not every repeated conversation. Keep a subscription’s usage pool separate from metered API spending and verify how any linked API access is charged. Budget from usage records rather than visible answer length.
How to use this data
- Recalculate with your real input/output token mix.
- Check region, rate limits, and payment method.
- Recheck official sources before purchasing or switching.
Frequently asked questions
Is a Grok subscription the same as API credit?
No. Subscription entitlements and metered API billing are different. Verify any explicitly included API access and its limits instead of treating the monthly fee as cash credit.
Are reasoning tokens charged when they are not visible?
Reasoning tokens are billable output. Use the reported output total and avoid adding reasoning a second time when it is already included.
Is live search included in the token rate?
Search and server-side tools have separate charges as well as associated token usage. Check whether the selected tool bills calls or retrieved items.
Can I apply cached-input pricing to all history?
Apply it only to input confirmed as cached in usage metrics. The remaining history uses the ordinary input rate for the selected model.
Official sources and review date
Billing guidance reviewed: . Price verification dates are listed separately in the table.