Grok 4.20 vs Grok 4.7

Best cell per row highlighted. Null means undisclosed — never counted as zero.

xAI
Grok 4.20GA
6.9
AI Panel
xAI
Grok 4.7GA
8.2
AI Panel

Identity & lifecycle

ProviderxAIxAI
Family / tierGrok 4 · ReasoningGrok 4 · Flagship
StatusGAGA
Released2026-03-102026-09-21
Knowledge cutoff2024-112026-05

Architecture & context

Context window1M500K
Max output tokens——
Input modalitiestext, imagetext, image
Reasoning modeoptionalalways
Open weightsNoNo

Pricing (per Mtok)

Input$1.25$2
Output$2.5$6
Cached input$0.2$0.5
Batch input$1—
Free tierYesYes

Speed

Output speed (tok/s)171.4—
Time to first token (s)13.24—
Latency tiermediummedium

Trust & deployment

Trains on inputsNoNo
LicenseProprietaryProprietary
Clouds——
ComplianceSOC2SOC2

AI Panel scoring

Unified score6.98.2
Decision Maker78
Domain Strategist6.58.5
Finance Lead6.59
Domain Practitioner7.58.5
Power User6.58
Skeptic67
Value score78.8

Benchmarks

GPQA Diamond78.5—
MATH-50087.3—
IFEval81—
TAU-bench93—
LMArena Elo1475—
LMArena Coding Elo1510—

Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.