Grok 4.20 vs Grok Build 0.1 vs Grok 4.3

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

xAI
Grok 4.20GA
6.9
AI Panel
xAI
Grok Build 0.1GA
6.7
AI Panel
xAI
Grok 4.3GA
7.7
AI Panel

Identity & lifecycle

ProviderxAIxAIxAI
Family / tierGrok 4 · ReasoningGrok Build · CoderGrok 4 · Flagship
StatusGAGAGA
Released2026-03-092026-05-192026-04-29
Knowledge cutoff2024-112024-112025-12

Architecture & context

Context window1M256K1M
Max output tokens256K
Input modalitiestext, imagetext, imagetext, image, video
Reasoning modeoptionalalwaysalways
Open weightsNoNoNo

Pricing (per Mtok)

Input$1.25$1$1.25
Output$2.5$2$2.5
Cached input$0.2$0.2$0.2
Batch input
Free tierYesNoYes

Speed

Output speed (tok/s)171.4181.5
Time to first token (s)13.2419.69
Latency tiermediumslowslow

Trust & deployment

Trains on inputsYesNoYes
LicenseProprietaryProprietaryProprietary
CloudsAzure AI Foundry
Compliance

AI Panel scoring

Unified score6.96.77.7
Decision Maker777.5
Domain Strategist6.578
Finance Lead6.57.58.5
Domain Practitioner7.577.5
Power User6.56.57.5
Skeptic666.5
Value score77.58.5

Benchmarks

GPQA Diamond78.590
MATH-50087.3
SWE-bench Verified70.8
IFEval8181
TAU-bench9398
LMArena Elo1491
Artificial Analysis Index4953

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