← Back to all models
Grok 4.6
xAILarge Language ModelProprietaryxAI's current flagship for coding and agentic tasks, released August 2026. 500K context, text and image input, reasoning effort levels from low to xhigh, and tiered pricing that doubles above 200K prompt tokens.
Abilities
Text GenerationCode GenerationReasoningMathVisionFunction CallingStructured OutputLong ContextMultilingual
Use Cases
Coding AssistantResearchData AnalysisChatbotAutomation
Available in Tools
Grokhttps://grok.com
Cursorhttps://cursor.com
Availability
How to Use
Use the xAI API with model ID `grok-4-6`. Reasoning effort is configurable (low/medium/high/xhigh, default high). Pricing is tiered: $2/$6 per MTok under 200K prompt tokens, $4/$12 above; cached input $0.50/$1.
Pros
- Strong coding and agentic performance — the current top of the Grok line
- 500K context window with no fixed output limit
- Configurable reasoning effort (low through xhigh) to trade cost against depth
- Competitive base pricing at $2/$6 per MTok with $0.50 cached input
- Recent knowledge cutoff (February 2026) plus real-time X data access in Grok products
Cons
- Pricing doubles above 200K prompt tokens ($4/$12) — long-context work costs more than headline rates
- Closed source — no self-hosting
- Rapid release cadence (4.5 in July, 4.6 in August) makes model pinning short-lived
- Smaller enterprise ecosystem than OpenAI, Anthropic, or Google
What to Use It For
Perfect For
Agentic coding workflows
Officially positioned as the flagship for coding/agentic tasks, with IDE integrations like Cursor
Current-events-aware assistance
Recent cutoff and real-time X integration in Grok products
Good For
Cost-tuned reasoning
Effort levels let you pay for depth only when tasks demand it
Not Recommended
Very long single prompts on a budget
The over-200K tier doubles cost — 1M-context competitors without tiering can be cheaper
Try instead: Claude Sonnet 5, Gemini 3.8 Flash
Self-hosting or offline use
Closed model — API only
Try instead: Kimi K3, DeepSeek-V4.1-Flash
Do Not Use For
Image or video generation
Text/vision LLM — use the separate Grok Imagine models for media generation
Try instead: Grok 4, FLUX.2 Pro
Technical Details
Pricing$2 / 1M input, $6 / 1M output (under 200K prompt tokens; $4 / $12 above; cached input $0.50)
ParametersUndisclosed
detail.contextWindow500K tokens