Grok 4 Models by xAI

xAIProprietary
xAI releases · 15 in the last 12 months · this family litChangelog →
16 models2025–2026Up to 2m ctxFrom $0.200/1M input

Last refreshed 2026-09-21. Next refresh: weekly.

Details

ResearcherxAI
LicenseProprietary
Commercial useCommercial use: conditional
Models16
Released2025–2026
Max context2m

Capabilities

Vision15 of 16 models
Multimodal15 of 16 models
Reasoning13 of 16 models
JSON / Tool use13 of 16 models
Structured Outputs11 of 16 models
Code Execution4 of 16 models

About

xAI's Grok 4 model family spans the original Grok 4 API lineup, Grok 4.3, Grok 4.20 variants, and Grok 4.5 — the July 2026 flagship for coding, agentic tasks, and knowledge work.

Decision facts

Best fit
agenticreasoningvision and multimodal work
Capability starting point
Grok 4.20 with 1m context and reasoning, JSON / Tool use, structured outputs, and multimodal inputs
Lowest tracked input
Grok 4.1 Fast · $0.200/1M · OpenRouter
Closest related family
Grok 1

Current Variants

Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.

11 in view5 retired
Grok 4.7Current

Use when the workload needs 500k context, reasoning, and JSON / Tool use.

2026-09500k contextreasoningJSON / Tool use
Grok 4.6Current

Use when the workload needs 500k context, reasoning, and JSON / Tool use.

2026-08500k contextreasoningJSON / Tool use
Grok 4.5Current

Use when the workload needs code, 500k context, and reasoning.

2026-07code500k contextreasoning

Use when the workload needs code, 256k context, and reasoning.

2026-05code256k contextreasoning
Grok 4.3Current

Use when the workload needs 1m context, reasoning, and JSON / Tool use.

2026-051m contextreasoningJSON / Tool use

Use when the workload needs agentic, 1m context, and reasoning.

2026-03agentic1m contextreasoning

Use when the workload needs 1m context, JSON / Tool use, and structured outputs.

2026-031m contextJSON / Tool usestructured outputs

Use when the workload needs reasoning, 1m context, and JSON / Tool use.

2026-03reasoning1m contextJSON / Tool use
Grok 4.20Current

Use when the workload needs 1m context, reasoning, and JSON / Tool use.

2026-021m contextreasoningJSON / Tool use
Grok 4.1Current

Use when the workload needs 131k context, reasoning, and JSON / Tool use.

2025-11131k contextreasoningJSON / Tool use

Use when the workload needs 256k context and multimodal inputs.

2025-07256k contextmultimodal inputs

Release Timeline

10 release groups
2026-09
1 current
Grok 4.7
500k contextreasoningJSON / Tool use
Current
2026-08
1 current
Grok 4.6
500k contextreasoningJSON / Tool use
Current
2026-07
1 current
Grok 4.5
code500k contextreasoning
Current
2026-05
2 current
Grok 4.3
1m contextreasoningJSON / Tool use
Current
Grok Build 0.1
code256k contextreasoning
Current
2026-03
3 current
Grok 4.20 Multi-Agent
agentic1m contextreasoning
Current
Grok 4.20 Non-Reasoning
1m contextJSON / Tool usestructured outputs
Current
Grok 4.20 Reasoning
reasoning1m contextJSON / Tool use
Current
2026-02
1 current
Grok 4.20
1m contextreasoningJSON / Tool use
Current
2025-11
1 current · 1 retired
Grok 4.1
131k contextreasoningJSON / Tool use
Current
Grok 4.1 Fast
2m contextreasoningJSON / Tool use
Replaced
2025-09
2 retired
Grok 4 Fast Non-Reasoning
2m contextreasoningmultimodal inputs
Replaced
Grok 4 Fast Reasoning
2m contextreasoningmultimodal inputs
Replaced

Replaced By

Keep for legacy integrations; evaluate Grok 4.3 before new work.

Keep for legacy integrations; evaluate Grok 4.3 before new work.

Keep for legacy integrations; evaluate Grok 4.3 before new work.

Keep for legacy integrations; evaluate Grok Build 0.1 before new work.

Keep for legacy integrations; evaluate Grok 4.3 before new work.

Specifications(16 models)

Grok 4 model specifications comparison
ModelReleasedContextParametersVisionMultimodalReasoningJSON / Tool useStructured OutputsCode Exec
Grok 4.72026-09500k—YesYesYesYesYesYes
Grok 4.62026-08500k—YesYesYesYesYesYes
Grok 4.52026-07500k—YesYesYesYesNoYes
Grok Build 0.12026-05256k—YesYesYesYesYesNo
Grok 4.32026-051m~0.5TYesYesYesYesYesNo
Grok 4.20 Multi-Agent2026-031m—YesYesYesYesYesNo
Grok 4.20 Non-Reasoning2026-031m—YesYesNoYesYesNo
Grok 4.20 Reasoning2026-031m—YesYesYesYesYesNo
Grok 4.202026-021m—YesYesYesYesYesNo
Grok 4.12025-11131k—YesYesYesYesNoNo
Grok 4 Heavy2025-07256k—YesYesNoNoNoNo

Pricing

Grok 4 model pricing by provider
ModelProviderInput / 1MOutput / 1MType
Grok Build 0.1xAI Console$1$2Serverless
Grok Build 0.1Vercel AI Gateway$1$2Serverless
Grok Build 0.1OpenRouter$1$2Serverless
Grok 4.20xAI Console$1.25$2.5Serverless
Grok 4.3xAI Console$1.25$2.5Serverless
Grok 4.20 Multi-AgentxAI Console$1.25$2.5Serverless
Grok 4.3OpenRouter$1.25$2.5Serverless
Grok 4.20OpenRouter$1.25$2.5Serverless
Grok 4.20 ReasoningxAI Console$1.25$2.5Serverless
Grok 4.20 Non-ReasoningxAI Console$1.25$2.5Serverless
Grok 4.20 Multi-AgentVercel AI Gateway$1.25$2.5Serverless
Grok 4.20 Non-ReasoningVercel AI Gateway$1.25$2.5Serverless
Grok 4.20 ReasoningVercel AI Gateway$1.25$2.5Serverless
Grok 4.3Vercel AI Gateway$1.25$2.5Serverless
Grok 4.3AWS Bedrock$1.25$2.5Serverless
Grok 4.7OpenRouter$1.6$4.8Serverless
Grok 4.20 Multi-AgentOpenRouter$2$6Serverless
Grok 4.5xAI Console$2$6Serverless
Grok 4.5OpenRouter$2$6Serverless
Grok 4.5Vercel AI Gateway$2$6Serverless
Grok 4.6xAI Console$2$6Serverless
Grok 4.6OpenRouter$2$6Serverless
Grok 4.7xAI Console$2$6Serverless

Popular comparisons in this family