Model
Model explorer

Grok-4

CLOSED
SpaceXAI · Grok 4 family · released Jul 9, 2025

Native tool-use reasoning flagship; led ARC-AGI-2 at launch. Official model card published by xAI (data.x.ai). Deprecated in the May 15, 2026 retirement wave, redirects to Grok 4.3.

ReasoningCodingVisionFunction callingTool useAgentic
1968.6
Elo · rank #61
Parameters
Undisclosed
Active params
Undisclosed
Context
256K tokens
Architecture
Mixture-of-Experts (size undisclosed; RL-trained at pretraining scale on Colossus, ~200K H100s)
License
Proprietary
Languages
—
API price (in/out)
$3 / $15
Modalities
text · vision
Benchmark results
Bar shows position within the tracked field; marker = field best
Aider PolyglotCoding79.6%#5
best: Claude Opus 4.5 · 89.4%
AIMEMath91.7%#38
best: GPT-5.2 · 100.0%
ARC-AGI-1Reasoning66.6%#10
best: GPT-5.6 · 96.5%
ARC-AGI-2Reasoning15.9%#22
best: GPT-5.6 · 92.5%
GPQA DiamondReasoning87.5%#41
best: GPT-5.6 · 94.6%
Humanity's Last ExamReasoning25.4%#53
best: Claude Sonnet 5 · 57.4%
LiveCodeBenchCoding79.0%#52
best: DeepSeek-V4-Pro (Think Max) · 93.5%
MMMUVision76.3%#33
best: Claude Fable 5 · 89.3%
SWE-bench VerifiedCoding57.8%#80
best: Claude Fable 5 · 95.0%
USAMOMath37.5%#5
best: Claude Opus 4.8 · 96.7%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$3
Output / M tok
$15
Grok 4 family
Elo progression across releases
Grok-4Jul 20251968.6Grok 4 HeavyJul 20252339.9
Resources
DOCSdocs.x.aiAPIAvailable via SpaceXAI API
API price $3/$15 · each benchmark row carries its own source badge (see methodology)