Xai
Grok 4.20 Multi-Agent 16
A multi-agent Grok variant configured for higher agent depth, designed for highly complex research tasks. The maximum output length is not explicitly capped and draws directly from the remaining 2M context window (note that internal agent reasoning also consumes this budget).
Quick facts
Pricing
Input $2 / Output $6 per 1M tokens
Context window
2M
Max output
2M
Knowledge cutoff
2024-11
What this model can do
You provide
- 📄 Documents (PDFs, Word, TXT)
- 🖼️ Images & Screenshots
- 💬 Text instructions
It creates
- 📝 Formatted text & summaries
- 📊 Excel sheets & CSV data
- 📽️ PowerPoint slides & decks
- 💻 Usable code snippets
- 🤖 JSON & structured data
Capabilities
Reasoning
Better suited to multi-step analysis, planning, and harder prompts.
Vision
Can work with screenshots, charts, or other image inputs.
Document input
Can accept supported documents directly in chat.
Structured output
Can return JSON for workflows that need machine-readable output.
Tool use
Can use compatible tools when crowSync enables them for the run.
Native tools in crowSync
Web search
Search the live web during a run.
Available
X search
Search X posts, profiles, and threads during a run.
Available
Code interpreter
Run code for calculations, analysis, and generated files.
Available
Input and output
Input
- Text
- Image
- Documents
Output
- Text
Common use cases
- Complex planning, debugging, and multi-step analysis.
- Research tasks that need current information from the web.
- Summarizing PDFs, documents, or longer working files.
- Data analysis, calculations, and code-backed workflows.
Pricing details
Input
$2 / 1M tokens
Output
$6 / 1M tokens
Cached input
$0.20 / 1M tokens