Batch summaries
100k input tokens and 50k output tokens for a content batch.
Grok calculator
Grok Token Calculator
Price a Grok workload with xAI's short-context text rates before tools, search calls, or priority processing are added.
Prices verified: via OpenRouter
Interactive tool
Convert tokens to cost
Use presets, share the exact inputs, and scan the live breakdown.
Estimated total cost$0.50
Cost per 1k tokens$0.0033
Examples
Short support reply
2,000 prompt tokens and a compact 300-token answer.
Grok 4.5: $2.00 input / $6.00 output per 1M tokens.
How this Grok Token Calculator works
How the current Grok tiers are priced
Grok 4.5 is xAI's premium general model and is listed at $2 input and $6 output per 1M tokens. Grok 4.3 is the lower-priced general tier at $1.25 input and $2.5 output per 1M tokens, while the mapped Grok 4.20 reasoning route uses $1.25 input and $2.5 output per 1M tokens. The two lower-cost entries share the same standard short-context rates, so the decision between them is driven by capability, reasoning behavior, and API compatibility rather than the baseline token bill. Grok 4.5 asks for a clear premium; benchmark it on the exact coding, knowledge, or agent tasks that justify that increase instead of treating the newest label as an automatic default.
Grok's input and output asymmetry
Grok 4.3 and 4.20 charge twice as much for output as for standard input. Grok 4.5 widens the gap: each output token is priced at three times an input token. An input-heavy classifier or document question can therefore have a very different unit cost from an agent that produces long plans, code patches, and explanations, even when both process the same total token count. The separate input and output fields above make that asymmetry visible. Include hidden reasoning or completion usage reported by xAI in the output side, and set realistic response limits rather than budgeting from prompt length alone.
Long context, caching, and xAI tools
xAI applies higher long-context rates once a qualifying prompt crosses its published threshold; this calculator uses the standard short-context prices in the central store. Cached input can cost less, whereas priority processing can apply a premium. Web Search, X Search, code execution, file handling, and other server-side tools are also billed outside plain text tokens. Those extras are especially relevant to Grok because tool-connected research is a common use case. Use this page for the model portion of the bill, then add observed tool invocations and check whether your typical prompt falls into xAI's long-context band before forecasting production spend.
When Grok is the economical or costly choice
Grok 4.3 or 4.20 can be the economical xAI route when you need configurable reasoning, strong tool calling, or X-native workflows without paying the 4.5 premium. Grok 4.5 becomes the expensive choice for output-heavy generation because both its base rate and output multiplier are higher. It earns its place only when better task completion, fewer corrective turns, or stronger coding performance lowers the cost of the entire workflow. For simple extraction and bulk transformation, cheaper sibling providers may win on tokens. For live research, compare total model-plus-tool cost and quality rather than token rates in isolation.
Examples
Grok 4.5 · 1,000,000 input and 500,000 output tokens. From the central pricing store ($2 input and $6 output per 1M tokens), input costs $2, output costs $3, and the total is $5.
FAQ
Does this include Grok Web Search or X Search charges?
No. The result covers text input and output tokens only. Add xAI's current per-invocation charge for each server-side search or code-execution tool your workflow uses.
Why can a long Grok prompt cost more than the calculator shows?
xAI publishes a higher rate band for qualifying long-context requests. This page uses standard short-context rates, so check the current threshold when prompts approach hundreds of thousands of tokens.
Is Grok 4.5 always more cost-effective than Grok 4.3?
No. Grok 4.5 costs more per input and output token. It is cost-effective only when its quality reduces retries, manual correction, or the number of model calls enough to offset that premium.