AI provider / xAI
xAI
A provider associated with Grok, real-time product experiences and a fast-moving frontier narrative around reasoning, coding and multimodal AI.
xAI is the fast-moving frontier lab behind Grok, built to fold real-time information and a more direct assistant voice into a single product. From Grok-1 — later open-sourced under Apache 2.0 — to GPT-4-class Grok-2, the line scaled quickly and publicly.
Grok 3 debuted on the 200K-GPU Colossus cluster, Grok 4 claimed top reasoning with native tool use, and the 4.x series added Fast and code-fast spinoffs plus multi-agent experiments. The defining trait is pace: an aggressive product cadence tied directly to live environments.
Model catalog
Every release in this provider’s model line.
Each row is a distinct model release — its date, its type and the traits it is known for. Open a source to dig into any of them.
| Model | Released | Type | Key traits | Source |
|---|---|---|---|---|
| Grok 4.7 | Sep 2026 | Flagship |
| |
| Grok 4.6 | Aug 2026 | Version update |
| |
| Grok 4.5 | Jul 2026 | Flagship |
| |
| Grok 4.20 Beta | Feb 2026 | Experimental |
| |
| Grok 4.1 | Nov 2025 | Version update |
| |
| Grok 4 | Jul 2025 | Flagship |
| |
| Grok 3 | Feb 2025 | Flagship |
| |
| Grok-2 | Aug 2024 | Version update |
| |
| Grok-1 (open weights) | Mar 2024 | Open release |
| |
| Grok-1 | Nov 2023 | Debut |
|
Timeline
A practical timeline of model releases and company milestones.
This page tracks the moments that matter most: major model releases, version jumps and company moves that changed how this provider shows up in the AI market.
Grok 4.7 runs on a 2.1-trillion-parameter base, a 40% jump over 4.6.
Trained with a longer reinforcement-learning run on harder, hours-long tasks, it works longer on difficult problems and checks its own work more carefully, with xAI’s best-calibrated safeguards yet — Musk now places 4.8 and 4.9 before Grok 5.
Grok 4.6 pushes toward long-running agents and matches GPT-5.6 Sol.
With a 500K-token context window and a focus on more ambitious interactive and visual work, 4.6 reaches frontier intelligence on the Artificial Analysis Index.
Grok 4.5 ships everywhere at once, from grok.com to X.
xAI’s smartest model for coding, agentic tasks and knowledge work rolls out simultaneously across web, X and the iOS and Android apps.
Grok 4.20 Beta experiments with multi-agent architecture.
A 4-agent system, then a 16-agent Heavy variant, preview the orchestration approach intended for Grok 5.
Grok 4.1 refines the flagship after Fast and code-fast spinoffs.
September’s Grok 4 Fast and the grok-code-fast coding line get folded into a steadier release rhythm.
Grok 4 claims top reasoning scores with native tool use.
A Heavy multi-agent tier and real-time search integration mark xAI’s shift from chasing to setting benchmarks.
Grok 3 debuts on the 200K-GPU Colossus cluster.
Trained with 10x Grok-2’s compute, it ships with Think mode and DeepSearch — scale as the headline argument.
Grok-2 reaches GPT-4 class and adds image generation.
After Grok-1.5 brought long context and vision in spring, Grok-2 closes the frontier gap in under a year.
Grok-1 weights go open under Apache 2.0.
The 314B release lands amid the Musk-OpenAI lawsuit, doubling as a statement about what “open” AI should mean.
Grok-1 launches with real-time X data and an attitude.
Four months after xAI’s founding, the 314B MoE debuts inside X Premium, differentiating on live data and irreverence.