Skip to main content

grok · xai · chat & research

Grok, for what's happening now.

xAI's models for quick takes, code and live search across the web and X, without a SuperGrok plan.

made by
xAI
versions
4
from
5 tokens
included on
Starter+

best at

What Grok is good at

Grok's personality is the first thing people notice: direct, brisk, and happy to give an opinion when asked for one. That makes it a good sparring partner for ideas and a quick way to get a straight answer without three paragraphs of preamble.

Its other strength is the present tense. Turn on Search and Grok pulls in real-time results from the web and from X, which suits questions about what people are saying, how a story is developing, or what changed today. Grok 4.3 accepts very long prompts too, up to a million tokens on the LazySusan API, so a long transcript or log file fits in one go.

For code there is Grok Build, a version tuned for programming. And the Image switch in the composer creates pictures with Grok Imagine without leaving the chat, which is handy for a quick visual to go with a post.

versions

Grok versions on LazySusan

Grok versions available on LazySusan
VersionUse it for
Grok AutoPicks per message: Grok Build for code questions, Grok 4.7 for why, explain, compare and longer prompts, Grok 4.3 for the rest.
Grok 4.7xAI's flagship reasoning model, for analysis and questions that need thinking through.
Grok 4.3Fast, with room for very long inputs. The everyday choice.
Grok BuildOptimised for code: writing, explaining and debugging.

pricing

What Grok costs on LazySusan

Grok token costs on LazySusan
GenerationTokens
Grok Build3–9 from a short to a long reply5
Grok 4.33–12 from a short to a long reply6
Grok 4.76–25 from a short to a long reply11
Grok Imagine, one image from the chat24

Chat costs are for a message of about 300 characters; you pay for what each reply uses. A typical Grok 4.3 reply is 6 tokens and Grok 4.7 11 tokens. Live search adds a little to each message that uses it.

Every plan includes Grok. Starter is $8 a month with 1,000 tokens; unused tokens roll over while you're subscribed. Compare plans.

prompts

Prompts to try with Grok

  • // live pulse

    With Search on: what are people on X saying about today's interest rate decision? Summarise the main camps and link a few representative posts.

  • // straight answer

    Give me your honest take: is it worth learning Rust in 2026 if I mainly write Python data pipelines? Keep it to five sentences.

  • // code

    This Python function times out on large inputs. Explain why and rewrite it to run in linear time.

  • // long input

    Here is the full server log from last night. Find the first error that caused the cascade and explain the chain of events.

  • // image

    With Image on: a retro travel poster for a fictional Martian ski resort, bold flat colours.

tips

Getting better results

  1. 01Ask Grok for an opinion when you want one. It is more willing than most models to commit to a view, which is useful for decisions and less useful for neutral summaries.
  2. 02Turn on Search for anything time-sensitive. Without it, Grok answers from its training data, which won't know about this morning.
  3. 03Grok Auto sends code questions to Grok Build, at about 5 tokens a reply. If you're doing a long coding session, select Grok Build directly to keep it there.
  4. 04Grok Imagine images from the chat are about 24 tokens each, and you can ask for several at once.

alternatives

When to pick a different tool

also on the api

Grok in your own app

Grok's chat models, and Grok Imagine for images, through the same key and SDK as every other model.

xai/grok-4.7xai/grok-4.3xai/grok-build-0.1

One key, our SDKs for JavaScript and Python, prepaid credits with no subscription. LazySusan API · docs

faq

Grok on LazySusan: questions

get started

Grok and every other model, one plan.

One login, one bill and one Library for everything you make, with the right model for each part of the job.

Starter is $8/month. Cancel anytime.