AgentHubAgentHub

AI tools·Chat models·Global

Grok

xAI’s flagship: realtime info, a sharp tone, and coding. Now in Cursor, GitHub Copilot, and Grok Build.

9.5/ 10.0Editorial Rating
Editorial score based on public benchmarks and community feedback — not a third-party benchmark result

Compiled through 2026-09

Grok visual
Overseas accessFree tierRealtime infoReasoningCodingCompare
Vendor
xAI
Started
2023
China availability
Overseas accessRequires overseas network access
Pricing tier
Free tier
Current flagship
Grok 4.6 / Grok 3
Website
grok.com

Evaluation, Pros & Cons

Editorial score based on public benchmarks and community feedback — not a third-party benchmark result

Key Advantages

  • Real-time access to X data stream with unmatched breaking news latency
  • Trained on massive Colossus cluster; major leaps in math and agentic coding

Limitations & Caveats

  • Direct/unfiltered style; requires X Premium or dedicated subscription

Overview

Grok is xAI’s model, originally tied to realtime X data. Grok 3 (2025) closed much of the quality gap. Grok 4.5 launched Jul 16, 2026 (trained with Cursor; coding and agents), then Grok 4.6 on Aug 12, quickly landing in GitHub Copilot across VS Code, CLI, and cloud agents. Priced below Opus / GPT-5.6 Sol for long coding jobs.

Direction

The path is a native chat app, a developer API, and distribution inside major IDEs. Grok Build targets end-to-end apps and spreadsheet modeling. Realtime search, multimodal, and longer agent jobs are next.

Version history

  1. Aug 2026

    Latest

    Grok 4.6

    Aug 12 launch; Copilot on Aug 14.

    View details

    New capabilities

    • Grok 4.6 on 12 August
    • Copilot surfaces on 14 August

    What got stronger

    • Moved from a Cursor option to a GitHub-workflow option

    Version notes

    4.6 was mostly distribution: Copilot’s VS Code, CLI, Cloud Agent, and more. You did not have to switch editors to use xAI.

    #2026-08-grok-4-6

  2. Jul 2026

    Grok 4.5

    Jul 16; coding agents; in Cursor.

    View details

    New capabilities

    • Trained with Cursor; coding agents as the pitch
    • Launched 16 July and landed in Cursor’s picker

    What got stronger

    • Long coding jobs became cost-competitive

    Version notes

    Grok 4.5 pivoted to engineering: trained with Cursor to work in the editor. Priced below Opus / GPT flagships for long agent runs.

    #2026-07-grok-4-5

  3. Feb 2025

    Grok 3

    Reasoning and coding join the frontier.

    View details

    New capabilities

    • Reasoning and coding entered the frontier conversation
    • Started getting compared with GPT / Claude on the same stage

    What got stronger

    • Code and math stopped being obvious weak spots

    Version notes

    Grok 3 was xAI proving it was more than personality. That made 4.5/4.6 in Cursor and Copilot a credible next step.

    #2025-02-grok-3

  4. Dec 2024

    Grok 2

    Image gen and better chat.

    View details

    New capabilities

    • Image generation in the product
    • Clearly better chat than v1

    What got stronger

    • Became a daily assistant, not only an X toy

    Version notes

    Grok 2 added multimodal and stabler chat. Realtime search stayed the differentiator; quality became good enough for mail and copy.

    #2024-12-grok-2

  5. Mar 2024

    Grok open weights

    Early open weights for the community.

    View details

    New capabilities

    • Early open weights for community fine-tunes

    What got stronger

    • Open weights complemented the closed chat app

    Version notes

    Open weights put Grok in researcher and local-deploy conversations, even though the product stayed grok.com / X. The goal was inspectability, not then-best chat.

    #2024-03-grok-open-weights

  6. Nov 2023

    Grok 1

    X Premium preview; realtime tone.

    View details

    New capabilities

    • Realtime feed tied to X
    • Sharper, more humorous tone

    What got stronger

    • Fresher on news and trends than closed-door models

    Version notes

    Grok 1 launched via X Premium. The pitch was realtime and personality, not then-best reasoning. It set xAI’s “assistant with an attitude” persona.

    #2023-11-grok-1