xAI: Grok 4.20 API
Grok-4.20-beta-0309-reasoning is the elite, high-intelligence tier of the Grok-4.20 family. It is built on a revolutionary "Society of Mind" multi-agent architecture, where four specialized agents (Captain, Research, Logic, and Critic) work in parallel to debate and verify findings before delivering the final response. Trained on the Colossus 200k GPU cluster using advanced Reinforcement Learning (RL), this model is specifically optimized for Expert-level tasks in medicine, legal analysis, and complex software architecture. It features "Chain-of-Thought" (CoT) reasoning that is 6x more compute-efficient than previous versions, allowing it to solve "Humanity's Last Exam" (HLE) level problems with unprecedented accuracy.
- Context window: 2,000,000 tokens
- Max output: 2,000,000 tokens
- Input: text, image
- Output: text
- Reasoning: Supported
- Tool calling: Supported
- File input: Supported
- Released: 2026-03-08
- Knowledge cutoff: 2026-02-28
Frequently Asked Questions
What is the context window of Grok 4.20?
Grok 4.20 supports a context window of up to 2,000,000 tokens.
Does Grok 4.20 support function calling?
Yes. Grok 4.20 supports tool / function calling.
Does Grok 4.20 support reasoning?
Yes. Grok 4.20 is a reasoning-capable model.
What is the knowledge cutoff of Grok 4.20?
The knowledge cutoff of Grok 4.20 is 2026-02-28.