Updated

A black obsidian shard split diagonally on a dark background, abstract hero artwork for Grok 4.6

xAI

Grok 4.6

xAI's Grok 4.6 for turning a broad product idea into a working first version, with 500K context.

Chat with Grok 4.6 in Super AgentView pricing

What is Grok 4.6?

Grok 4.6 builds on Grok 4.5 with a focus on long-running agents and more ambitious interactive and visual work.

xAI released it on August 12, 2026 and says it is especially strong at turning a broad product idea into a working first version, testing and verifying more of its own work on long tasks. Its first passes on visual and interactive projects are stronger than Grok 4.5's. It reads text and images with a 500K-token context window, and xAI has since released Grok 4.7 as its successor.

Provider
xAI
Context window
500K tokens
Input
Text, Image
Reasoning
Low to extra high
Knowledge cutoff
January 2026
Released
Aug 12, 2026

How Grok 4.6 performs

Published by xAI

Results xAI reported for Grok 4.6. Each lab runs its own setup, so compare these with other models only inside the same table.

Benchmark results xAI published for Grok 4.6
BenchmarkScoreSetting
APEX-SWE56.4%high effort
FrontierCode v1.161.3%high effort
DeepSWE v1.167%extra high effort
SWE-Marathon v1.131.9%high effort
Terminal-Bench 3.026%high effort
APEX-Agents57.5%high effort
OfficeQA Pro63.2%high effort
EEBench60%extra high effort
CADBench88.4%extra high effort
Hallucination rate (lower is better)1.7%high effort, xAI internal test

Source: media.x.ai

What you can do with Grok 4.6

  • Turns a broad product idea into a working first version, according to xAI

  • More self-testing and verification on long tasks, according to xAI

  • Stronger first passes on visual and interactive projects

  • Holds up to 500K tokens of text and images

How to prompt Grok 4.6

  • Start with the broad idea and let it propose a first version before you add detail.
  • Use extra high effort for the hardest steps.
  • Run the same prompt on Grok 4.7 to see what the newer version changes.

Where Grok 4.6 fits best

  • Prototyping an app or tool from a rough idea

  • Planning interactive and visual projects

  • Long agent runs that need self-checks

  • Comparing against Grok 4.7

How Grok 4.6 compares

Grok 4.6 compared with other chat models on Flashloop
ModelLMArena textLMArena codingContext windowReasoning
Grok 4.61456#661510#60500K tokensLow to extra high
Gemini 3.5 Flash Lite1455#711502#771M tokensLow, medium, high
DeepSeek V4 Pro1458#641501#791M tokensOff, medium, high, extra high
GPT-6 Sol1459#621519#381.05M tokensOff to max
GPT 5.6 Luna1453#751502#761.05M tokensOff to max

Good to know before you start

  • Grok 4.7 is its successor
  • No max reasoning setting; extra high is the top level
  • Grok 4.6 itself returns text; images and video in Super Agent come from Flashloop's generation models

Start chatting with Grok 4.6

Chat with Grok 4.6 in Super Agent

  1. Select Grok 4.6

    Open Super Agent on Flashloop and choose Grok 4.6 in the model picker.

  2. Enter your prompt

    Ask for what you need: an answer, a plan, or the images and videos you want the agent to make.

  3. Send and keep going

    Send the message. Grok 4.6 answers in Super Agent and runs any generations you asked for.

Questions about Grok 4.6

Is Grok 4.6 xAI's latest model?

No. xAI has released Grok 4.7 as its successor. Both are available in Super Agent.

What is Grok 4.6 good at?

Long-running agent work, which xAI made a focus of this release, and turning a broad product idea into a working first version, where xAI says it is especially strong. Its first passes on visual and interactive projects are also stronger than Grok 4.5's.

Which reasoning levels does Grok 4.6 support?

Low, medium, high and extra high. Super Agent starts it at low.

What is Grok 4.6's context window?

500K tokens, with text and image input.

More chat models on Flashloop

Browse all models