Grok Models Overview
On this page
Updated November 3, 2026 based on the Models page and xAI release notes: Grok 4.6 (August 12, 2026) and Grok 4.7 (September 21, 2026) replaced Grok 4.5 as the flagship models, featuring 500,000 context tokens, reasoning effort up to
xhigh, and the same tiered pricing; the table and history reflect these changes. Sources: Models · Release notes.
The Grok Lineup: A Complete Model Family
xAI, the company founded by Elon Musk, now offers a family of artificial intelligence models under the Grok brand. Far from being limited to a single model, this lineup covers a wide range of needs: advanced reasoning, fast generation, code, images, videos, and even real-time voice interaction.
If you work with AI APIs or want to integrate a high-performance model into your applications, understanding the range of Grok models is essential to making the right choice.
The xAI Philosophy
xAI stands out due to several key technical choices. First, the emphasis on reasoning: flagship models, led by Grok 4.7, offer adjustable reasoning—you adjust the thinking effort based on the task, instead of choosing between two distinct models. The grok-4.20 generation, however, retains separate reasoning and non-reasoning variants.
Next, pricing works in tiers: each model has a base rate as long as the request stays under 200,000 context tokens, and a doubled rate beyond that. Input tokens already in cache are billed at a significantly reduced rate, which lowers costs for applications that reuse the same prompt prefix.
Finally, native integration with the X platform offers unique search tools: web search and X search directly within API calls. Note that the x.ai site now displays the “SpaceXAI” brand, but the technical documentation remains on x.ai—we keep “xAI” in this course.
Evolution: From Grok 3 to Grok 4.7
The recent history of Grok models reads like a constant acceleration:
- April 2025 — Grok 3 becomes generally available via the API. This is the first model to establish xAI as a serious player in the market.
- July 2025 — Grok 4 arrives with enhanced internal reasoning and significantly better performance on complex tasks.
- March 2026 — Grok 4.20 is unveiled in three variants (reasoning, non-reasoning, and multi-agent), with a 1 million token context.
- June 1, 2026 — grok-build-0.1, the fastest code model in the lineup, enters public beta on the API.
- July 16, 2026 — Grok 4.5 opens the current flagship tier: 500,000 context tokens and adjustable reasoning. It deploys on iOS, Android, web, and X on July 22.
- August 12, 2026 — Grok 4.6 succeeds it, with
xhighreasoning effort. - September 21, 2026 — Grok 4.7 becomes the model xAI recommends for code, agentic tasks, and knowledge work.
Grok 3 (all variants, including grok-3-mini) and Grok 4 without a suffix belong to previous generations: do not use them for new projects.
The Current Lineup at a Glance
| Model | Specialty | Context | Input/Output Price per M (< 200k) |
|---|---|---|---|
| grok-4.7 | Model recommended by xAI, adjustable reasoning up to xhigh | 500k tokens | $2.00 / $6.00 |
| grok-4.6 | August 2026 generation, same pricing | 500k tokens | $2.00 / $6.00 |
| grok-4.5 | July 2026 generation, still in catalog | 500k tokens | $2.00 / $6.00 |
| grok-4.3 | Cost-effective versatile, very long context | 1M tokens | $1.25 / $2.50 |
| grok-4.20-0309 (reasoning / non-reasoning) | 4.20 generation, still in catalog | 1M tokens | $1.25 / $2.50 |
| grok-4.20-multi-agent-0309 | Agent orchestration | 1M tokens | $1.25 / $2.50 |
| grok-build-0.1 | Code & development, fastest | 256k tokens | $1.00 / $2.00 |
| Grok Imagine | Images & videos | — | Per unit / per second |
| Voice API | Voice agent, TTS, STT | — | Per hour / per character |
Beyond 200,000 context tokens, text model prices double: grok-4.7, for example, goes to $4.00 input and $12.00 output per million tokens.
What You Will Learn in This Course
This course will guide you through each model in the Grok lineup. You will discover the technical specifics of each one, learn to compare their pricing, and know how to choose the right model for your use case.
The lessons are organized into four progressive sections:
- Model Catalog — detailed presentation of each text and code model
- Multimedia Models — image generation, video generation, and voice interaction
- Pricing — understand costs and optimize them
- Choosing Your Model — decision method adapted to your project
Pricing verified November 3, 2026 — prices change regularly: before any budget calculation, check the current grid on the official xAI models and pricing page.
Key Takeaways
- The lineup includes eight text and code models (grok-4.7, grok-4.6, grok-4.5, grok-4.3, the three grok-4.20-0309 variants, grok-build-0.1), plus image, video, and voice
- grok-4.7 is the model xAI recommends (500k context, adjustable reasoning); grok-4.3 and grok-4.20 go up to 1 million tokens
- Pricing is tiered: base rate under 200k context tokens, doubled rate beyond
- Cached input tokens are billed at a reduced rate—structure your prompts to take advantage of this
- Grok 3 and Grok 4 without a suffix are previous generations: do not choose them for new projects