Grok Token Counter
Estimate your text in Grok tokens and price it across all 4 xAI models. Grok has the most consequential billing rule of any provider here: from 200,000 prompt tokens, xAI bills the whole request at double the listed input, cached and output rates. Knowing where your prompt sits relative to that line is worth more than any model choice.
- Countingest.
- Context256K–1M
- Input$2.00
- Output$6.00
- Cached$0.50025% of input
Estimate Grok tokens
Opens on Grok 4.6. Watch the count against 200,000 — that is the threshold where xAI re-prices the entire request, not just the tokens above it.
Drop files here or click to browse
Supports PDF, TXT, MD, JSON, CSV, XML — max 10MB
Count as an API request
Providers also bill the role and boundary tokens that wrap your text. Anthropic's own example — a short system prompt plus one message — costs 14 input tokens for 7 tokens of visible text.
Is this Grok token counter exact?
No. xAI exposes a /v1/tokenize-text endpoint that will tokenize a string for you, but it does not publish the vocabulary those tokens come from, so the segmentation cannot be reproduced offline. The count here is an estimate calibrated on English prose — close enough to tell whether you are near the 200,000-token pricing cliff, not close enough to reconcile an invoice line by line.
If you need the exact number, call /v1/tokenize-text with the text you plan to send. It is the authoritative count and it is what xAI bills against. Remember that it measures the string you pass, while a real request also carries a system prompt, tool definitions and any conversation history — and on Grok that overhead is not just extra tokens, it is what can push a request over the threshold and double the price of everything in it.
Grok models and prices
4 xAI models. The lowest rates in the family belong to Grok Build 0.1 at $1.00 in and $2.00 out, but it also has the smallest window (256K); Grok 4.3 costs a little more at $1.25 / $2.50 and holds 1M — the largest window xAI offers. Cheapest and roomiest are not the same row here.
| Model | Inputper 1M tokens | Cachedper 1M tokens | Outputper 1M tokens | Context | Counting | Your cost |
|---|---|---|---|---|---|---|
| Grok 4.6 | $2.00 | $0.500 | $6.00 | 500K | est. | — |
| Grok 4.5 | $2.00 | $0.300 | $6.00 | 500K | est. | — |
| Grok 4.3 | $1.25 | $0.200 | $2.50 | 1M | est. | — |
| Grok Build 0.1 | $1.00 | $0.200 | $2.00 | 256K | est. | — |
Rates in USD per 1M tokens, read from xAI’s own pricing page on August 22, 2026. Providers change prices without notice.
Grok’s context window, and the 200K cliff
Grok 4.6 accepts 500,000 tokens — about 746 pages of prose at roughly 670 tokens a page. Grok 4.3 goes further, to 1,000,000 tokens (1,493 pages), while Grok Build 0.1 stops at 256K.
Read those windows together with the pricing tier and they change meaning. Because xAI doubles the rate on the entire request from 200,000 prompt tokens, most of Grok 4.6’s window — 300,000 of its 500,000 tokens — is only reachable at twice the listed price. A 150,000-token prompt costs $0.300; a 250,000-token prompt costs $1.00 rather than the $0.500 the headline rate implies. Trimming a prompt from just over the line to just under it is the single largest saving available on this provider.
What Grok tokens cost
Three request shapes priced on Grok 4.6 and Grok 4.3 at the standard tier. The third row crosses 200,000 input tokens, so a real bill for it would be double what the table shows.
| Request | Grok 4.6 | Grok 4.3 |
|---|---|---|
| One chat turn1,000 in / 500 out | $0.0050 | $0.0025 |
| Under the tier line100,000 in / 2,000 out | $0.212 | $0.130 |
| Over the tier line1,000,000 in / 50,000 out | $2.30 | $1.38 |
Grok’s cache discounts are shallower than the industry norm and they vary by model. Where Anthropic and Google read cached input at 10% of their input rate on every model they sell, Grok 4.6 charges $0.500 against $2.00 — 25% — and the rest of the family lands between that and 15%. Caching is still worth configuring, but it will not carry a workload the way it does elsewhere.
Grok token counter FAQ
How many tokens is 1,000 words in Grok?
About 1,330 tokens for English prose, and 1,000 tokens is roughly 752 words. xAI publishes no vocabulary, so this is an estimate rather than a measurement — usable for sizing a prompt, and worth confirming with /v1/tokenize-text if the number decides whether you cross a pricing threshold.
What happens after 200,000 prompt tokens on Grok?
The price of the whole request doubles — input, cached and output alike — not just the tokens above the line. This is the most expensive detail in xAI’s pricing and the easiest to miss. On Grok 4.6 it means a 250,000-token prompt costs $1.00 instead of $0.500. Because Grok 4.6’s window is 500,000 tokens, 300,000 of it is only usable at the doubled rate.
Is this Grok token counter exact?
No. xAI offers a /v1/tokenize-text endpoint but keeps the underlying vocabulary private, so an offline counter cannot reproduce its output. Expect roughly 10–15% accuracy on English prose and less on code or structured data. Call /v1/tokenize-text for the authoritative figure.
What is Grok 4.6’s context window?
500,000 tokens, about 746 pages of prose. Grok 4.3 is the roomiest model xAI sells at 1,000,000 tokens (1,493 pages), and Grok Build 0.1 the tightest at 256K. Bear in mind that on every one of them the 200,000-token mark changes the price, so the usable window and the affordable window are different sizes.
How much does Grok 4.6 cost?
$2.00 per 1M input tokens, $6.00 per 1M output and $0.500 for cached reads, at the standard tier. That output rate is only three times input — tighter than most providers, several of which charge five or six times — so Grok is comparatively kind to generation-heavy workloads, as long as the prompt stays under 200,000 tokens.
Which Grok model is cheapest?
Grok Build 0.1, at $1.00 in and $2.00 out — below Grok 4.3’s $1.25 / $2.50 on both rates. The trade-off is room: it holds 256K against Grok 4.3’s 1M. For short, high-volume calls it is the cheaper row; for anything long, the larger window is worth the difference.
Grok against the field
xAI is the only provider here that doubles the rate on an entire request past a prompt threshold, which makes a like-for-like comparison harder than the headline rates suggest. Compare 56 models from 11 providers with those caveats spelled out, or open another brand’s counter below.
New to tokens? Read how tokenization works