Skip to content
Updated minutes ago
xAI

Grok 3

xAI · moe · 600B parameters · 131,072 context

Quality
90.0

Parameters

600B

Context Window

128K tokens

Architecture

MoE

Best GPU

B200 NVL (pair)

Cheapest API

$15.00/M

Quality Score

90/100

Intelligence Brief

Grok 3 is a 600B parameter Mixture-of-Experts (16 experts, 2 active) model from xAI, featuring Grouped Query Attention (GQA) with 96 layers and 12,288 hidden dimensions. With a 131,072 token context window, it supports tools, vision, structured output, code, math, multilingual, reasoning. On standardized benchmarks, it achieves MMLU 89, HumanEval 70, GSM8K 95. The most cost-effective API deployment is via xai at $15.00/M output tokens. For self-hosted inference, B200 NVL (pair) delivers optimal throughput at $39858/month.

Architecture Details

TypeMOE
Total Parameters600B
Active Parameters120B
Layers96
Hidden Dimension12,288
Attention Heads96
KV Heads16
Head Dimension128
Vocab Size131,072
Total Experts16
Active Experts2

Memory Requirements

BF16 Weights

1200.0 GB

FP8 Weights

600.0 GB

INT4 Weights

300.0 GB

KV-Cache per Token2359296 bytes
Activation Estimate10.00 GB

Fits on (single GPU) — most practical first

GPU Compatibility Matrix

Grok 3 is compatible with 1% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
B300288GB
B100 SXM192GB
GB200 NVL72 (per GPU)192GB
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
H200 SXM141GB
H2096GB
GH20096GB
Ada Lovelace(11 GPUs)
L40S48GB
L4048GB
RTX 6000 Ada48GB
L2048GB
Ampere(16 GPUs)
A100 80GB SXM80GB
A100 80GB PCIe80GB
A1664GB
RTX A600048GB
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

B200 NVL (pair)good

BF16 · 4 GPUs · tensorrt-llm

68/100

score

Throughput

140.0 tok/s

Latency (ITL)

7.1ms

Est. TTFT

1ms

Cost/Month

$39858

Cost/M Tokens

$108.33

Use this config →
Instinct MI325Xgood

BF16 · 8 GPUs · vllm

65/100

score

Throughput

140.0 tok/s

Latency (ITL)

7.1ms

Est. TTFT

1ms

Cost/Month

$18904

Cost/M Tokens

$51.38

Use this config →
B200 SXMgood

BF16 · 8 GPUs · tensorrt-llm

63/100

score

Throughput

140.0 tok/s

Latency (ITL)

7.1ms

Est. TTFT

1ms

Cost/Month

$34088

Cost/M Tokens

$92.65

Use this config →

Deployment Options

API

API Deployment

xai

$15.00/M

output tokens

Self-Hosted

Single GPU

Requires multi-GPU setup (600 GB VRAM needed)

Scale

Multi-GPU

B200 NVL (pair) x4

140.0 tok/s

TP· $39858/mo

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
xai$3.00$15.00
Cheapest

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
xaiBest Value$3.00$15.00$90

Cost per 1,000 Requests

Short (500 tok)

$4.50

via xai

Medium (2K tok)

$18.00

via xai

Long (8K tok)

$54.00

via xai

Performance Estimates

Throughput by GPU

B200 NVL (pair)
140.0 tok/s
Instinct MI325X
140.0 tok/s
B200 SXM
140.0 tok/s

VRAM Breakdown (B200 NVL (pair), BF16)

Weights
Act
Weights 300.0 GBKV-Cache 12.9 GBActivations 80.0 GBOverhead 15.0 GB

Quality Benchmarks

Top 10%
98th percentile across all models
MMLU
89.0
Above Average (89th pctile)
HumanEval
70.0
Above Average (84th pctile)
GSM8K
95.0
Above Average (87th pctile)
MT-Bench
89.0
Bottom 25% (0th pctile)

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

Supported Precisions

BF16 (default)

Where to Deploy Grok 3

Similar Models

Grok-2

314B params · moe

Quality: 78

from $10.00/M

Lower quality, CheaperCompare →

Grok-3

314B params · dense

Quality: 91

from $15.00/M

Similar specsCompare →

Gemini 2.0 Pro

600B params · moe

Quality: 88

from $4.00/M

Larger context, CheaperCompare →
Smaller context, Lower qualityCompare →

DeepSeek R1

671B params · moe

Quality: 88

from $2.19/M

Frequently Asked Questions

How much VRAM does Grok 3 need for inference?

Grok 3 requires approximately 1200.0 GB of VRAM at BF16 precision, 600.0 GB at FP8, or 300.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (2359296 bytes per token) and activations (~10.00 GB).

What is the best GPU for Grok 3?

The top recommended GPU for Grok 3 is the B200 NVL (pair) (x4) using BF16 precision. It achieves approximately 140.0 tokens/sec at an estimated cost of $39858/month ($108.33/M tokens). Score: 68/100.

How much does Grok 3 inference cost?

Grok 3 API inference starts from $3.00/M input tokens and $15.00/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.