Grok

xAI: grok-4.20-0309-reasoning

grok-4.20-0309-reasoning
1M context2M outReasoningToolsVisionFilesCacheStructuredWeb search
Released Mar 31, 2026Knowledge cutoff 2025Updated Mar 31, 2026

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.

Mode chatTokenizer Grok

All providers for this model are busy right now

Every upstream provider has hit its rate limit. The model comes back automatically once limits lift, usually within hours. Try again in a little while or switch to another model.

Request this model on Discord

Pricing

Input price
$2.10/ 1M tokens
Output price
$4.20/ 1M tokens
Compatible endpoints anthropic, openaiVendor xAI

Performance

Loading performance data...

Usage & Ranking

Loading usage...

Supported parameters

All providers = every upstream serving this model supports it. Some providers = depends on which upstream handles the request. Default = the value sent when you leave the parameter unset.

ParameterProvidersDefault
frequency_penaltySome providersNot sent by default
include_reasoningAll providers-
logprobsAll providers-
max_tokensAll providers-
presence_penaltySome providersNot sent by default
reasoningAll providers-
repetition_penaltySome providersNot sent by default
response_formatAll providers-
seedAll providers-
structured_outputsAll providers-
temperatureAll providersNot sent by default
tool_choiceAll providers-
toolsAll providers-
top_kSome providersNot sent by default
top_logprobsAll providers-
top_pAll providersNot sent by default

Frequently asked questions

Similar models