Grok is xAI's frontier LLM family, exposed through an OpenAI-compatible API with a very large context window and live access to X data. Its headline is cost: the Grok 4.1 Fast tier is one of the cheapest frontier-class models on the market. Here's the pricing, the free credits, and how to get a key.
| Model | Input / 1M | Output / 1M |
|---|---|---|
| Grok 4.7 under 200K prompt | $2.00 | $6.00 |
| Grok 4.7 at/over 200K prompt | $4.00 | $12.00 |
| Grok 4.6 under 200K prompt | $2.00 | $6.00 |
| Grok 4.1 Fast cheapest | ~$0.20 | ~$0.50 |
| Grok 3 mini | ~$0.30 | ~$0.50 |
| Live Search (X data) | add-on, billed per source/request | |
Grok 4.6 and 4.7 both use a 200,000-token prompt cliff: cross that line and the entire request — input, cached input and output — bills at 2x, not just the tokens over the line. Cached input is $0.50/1M under the line, $1.00/1M over it. Both models are also the only ones on xAI's US regional endpoint, billed at a flat 1.1x premium on top of whichever tier applies. Full breakdown: the 200K-token cliff, priced.
xAI has typically offered promotional API credits rather than a permanent free tier — often in exchange for opting into data sharing — so the exact deal moves around. Grok chat is also included with X Premium, but that's separate from API access, which draws down a billed balance in your console. If you only need to experiment, claim whatever credit promotion is live; for production, budget per-token from the table above and price it with the calculators below.
1. Go to console.x.ai and sign in with your X account.
2. Add a payment method or claim available free credits.
3. Open API Keys and create a new key — copy it once (it won't be shown again).
4. Because the API is OpenAI-compatible, just change the base URL.
Call it with the OpenAI SDK shape:
Grok 4.1 Fast at roughly $0.20 / $0.50 per million tokens undercuts most frontier options — cheaper on output than GPT-class mid models and far below Claude Opus. The flagship Grok 4.7 sits at $2 / $6 per million for prompts under 200,000 tokens, in the same band as GPT-5-class models, but doubles to $4 / $12 the moment a prompt crosses that line — a pricing shape none of the other frontier APIs use. If cost is the priority, Grok Fast competes with Gemini Flash and DeepSeek; if you need its X-data and reasoning, budget for the 200K cliff explicitly. See the OpenAI, Claude, Gemini and DeepSeek guides, then rank them for your exact token mix with the cheapest LLM API tool.
Grok 4.6 and 4.7 bill $2 / $6 per million tokens for prompts under 200,000 tokens, doubling to $4 / $12 at or over that line (the whole request, not just the overage). The cheaper Grok 4.1 Fast is ~$0.20 / $0.50 (reference, September 2026). Confirm in the xAI console.
Yes — it follows the OpenAI chat-completions shape, so you can reuse the OpenAI SDK and just point the base URL at api.x.ai/v1 and swap the model name.
Usually promotional credits rather than a permanent free tier (sometimes tied to data sharing). X Premium includes Grok chat but not API access. Check the console for the current offer.
A very large context window, strong reasoning modes, and optional Live Search over X (Twitter) data — useful for real-time and social-aware applications.
Not affiliated with xAI. Prices are reference estimates — always verify on the official pricing page.