xAI API: models, pricing and limits
A practical guide to xai, using data from the AI & LLM Index. The goal is a shortlist and a decision, not a tour of every option.
The xAI lineup
xAI has 14 models in the index. AI company behind the Grok models. The lineup spans 3 categories, from frontier to mid-range.
The strongest model by average score is grok-4-reasoning. The cheapest by input price is Grok 3 Nano.
Pricing
Across the lineup, the average input price is $15.40 per million tokens and the average output price is $71.89. That average hides a wide range: individual models run from $7.12 to $19.70 on input.
Use the average as a starting point, then check the specific model you plan to deploy. Output price and cached input pricing can change the ranking for long or repetitive workloads.
Strengths and limits
0 of the 14 models are open weights. That gives you a self-hosting path for part of the lineup, while the proprietary models remain API-only.
The best benchmark result in the lineup is 87.8 on the index average. Compare that with the best model from other providers before treating xAI as the default.
How to choose
- Start with the task, not the brand.
- Pick the cheapest model that clears your evaluation set.
- Keep a second provider as a fallback for outages and price changes.
- Check licence and data-handling terms for every model you shortlist.
xAI models at a glance
| Model | Input / 1M | Context | Avg score |
|---|---|---|---|
| Grok 4 Reasoning | $19.67 | 2,000,000 | 87.8 |
| Grok 4 Turbo | $19.21 | 1,000,000 | 86.7 |
| Grok 4 Max | $19.70 | 2,000,000 | 86.6 |
| Grok 4 | $17.90 | 200,000 | 86.5 |
| Grok 3 Max | $17.66 | 2,000,000 | 85.9 |
| Grok 4 Pro | $19.00 | 1,000,000 | 85.4 |
| Grok 3 Pro | $16.61 | 400,000 | 83.8 |
| Grok 3 Turbo | $15.97 | 128,000 | 83.1 |
| Grok 3 Reasoning | $16.07 | 200,000 | 82.9 |
| Grok 3 | $15.84 | 128,000 | 82.8 |
Related
FAQ
Which xAI model should I start with?
Start with grok-4-reasoning for quality-sensitive work and Grok 3 Nano for high-volume tasks.
What is the most important factor in xai?
Fit to the task beats raw benchmark rank. Define the failure modes, pick the benchmarks that resemble them, then test a small shortlist on your own examples.
How many models should I evaluate?
Three is usually enough: a quality candidate, a value candidate and a fallback. More candidates add work without changing the decision.
How do I control cost?
Compare blended 3:1 price, watch output tokens, use caching where available and route simple requests to a cheaper model.