xAI ships several Grok models at the same time and the names do not always tell you which is which. The simplest way to think about the line-up is a Fast tier for volume work and a set of full-size models for harder reasoning. The chart below uses live prices from our catalogue.
The Fast vs full-size split
Grok 4 Fast is the budget option. At the time of writing our catalogue lists it at $0.20 per 1M input tokens and $0.50 per 1M output tokens. Grok 4 is the full-size model at $3 input and $15 output. That is a 15x gap on input and a 30x gap on output for the same conversation.
Between those two sit the newer numbered releases. Grok 4.3 and Grok 4.20 list at $1.25 input and $2.50 output. Grok 4.5 and Grok 4.6 list at $2 input and $6 output. Grok Build 0.1, a coding-oriented model, lists at $1 input and $2 output. Those figures come straight from models.json and update when xAI changes them.
| Model | Input | Output | Context window | Best for |
|---|---|---|---|---|
| Grok 4 Fast | $0.20 | $0.50 | Not listed | Volume: classification, extraction, summaries |
| Grok Build 0.1 | $1.00 | $2.00 | 256,000 | Coding tasks and agentic build loops |
| Grok 4.3 | $1.25 | $2.50 | 1,000,000 | General work with long inputs |
| Grok 4.20 | $1.25 | $2.50 | 2,000,000 | Whole-repo or multi-document jobs |
| Grok 4.5 | $2.00 | $6.00 | 500,000 | Harder reasoning at mid price |
| Grok 4.6 | $2.00 | $6.00 | 500,000 | Latest mid-tier release |
| Grok 4 | $3.00 | $15.00 | Not listed | Legacy flagship; usually not worth the premium |
Context windows: big numbers, real costs
Grok's headline feature is context. Our catalogue lists Grok 4.20 at 2,000,000 tokens, Grok 4.3 at 1,000,000, Grok 4.5 and Grok 4.6 at 500,000, and Grok Build 0.1 at 256,000. Grok 4 and Grok 4 Fast do not have a context window listed in our data, so check the official docs before relying on a figure.
A large window is a capability, not a free lunch. Every token you put in the window is billed as input. Filling Grok 4.20's 2M window costs about $2.50 per call at $1.25 per 1M. Do that fifty times in a session and you have spent $125 on reading alone. Send only what the model needs.
Which model when
- Bulk text processing (tagging, sentiment, extraction, short summaries): Grok 4 Fast.
- Reading long documents or code and answering questions about them: Grok 4.3 or Grok 4.20, depending on how much you need to load.
- Code generation and iteration: Grok Build 0.1 first, then Grok 4.6 if quality is not enough.
- Hard reasoning, careful analysis, high-stakes drafts: Grok 4.5 or Grok 4.6.
- Grok 4: only if you have a specific reason. It is the most expensive row in the catalogue and newer models are cheaper.
Knowledge check
At the prices in our catalogue, roughly how much more does Grok 4 charge for output tokens than Grok 4 Fast?