Skip to content
Updated minutes ago
Meta

Llama 4 Scout

Meta · moe · 109B parameters · 10,485,760 context

Quality
73.0

Parameters

109B

Context Window

10240K tokens

Architecture

MoE

Best GPU

B200 SXM

Cheapest API

$0.30/M

Quality Score

73/100

Intelligence Brief

Llama 4 Scout is a 109B parameter Mixture-of-Experts (16 experts, 1 active) model from Meta, featuring Grouped Query Attention (GQA) with 48 layers and 5,120 hidden dimensions. With a 10,485,760 token context window, it supports tools, vision, structured output, code, math, multilingual. On standardized benchmarks, it achieves MMLU 79, HumanEval 55, GSM8K 85. The most cost-effective API deployment is via together at $0.30/M output tokens. For self-hosted inference, B200 SXM delivers optimal throughput at $4261/month.

Architecture Details

TypeMOE
Total Parameters109B
Active Parameters17B
Layers48
Hidden Dimension5,120
Attention Heads40
KV Heads8
Head Dimension128
Vocab Size202,048
Total Experts16
Active Experts1

Memory Requirements

BF16 Weights

218.0 GB

FP8 Weights

109.0 GB

INT4 Weights

54.5 GB

KV-Cache per Token196608 bytes
Activation Estimate2.00 GB

GPU Compatibility Matrix

Llama 4 Scout is compatible with 21% 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 SXMoptimal

FP8 · 1 GPU · tensorrt-llm

100/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$4261

Cost/M Tokens

$5.79

Use this config →
B100 SXMoptimal

FP8 · 1 GPU · tensorrt-llm

100/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$4271

Cost/M Tokens

$5.80

Use this config →
GB200 NVL72 (per GPU)optimal

FP8 · 1 GPU · tensorrt-llm

100/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$6169

Cost/M Tokens

$8.38

Use this config →

Deployment Options

API

API Deployment

together

$0.30/M

output tokens

Self-Hosted

Single GPU

B200 SXM

$4261/mo

Min VRAM: 109 GB

Scale

Multi-GPU

H100 SXM x2

280.0 tok/s

TP· $3587/mo

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
together$0.18$0.30
Cheapest
fireworks$0.20$0.35

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
togetherBest Value$0.18$0.30$2
fireworks$0.20$0.35$3

Cost per 1,000 Requests

Short (500 tok)

$0.15

via together

Medium (2K tok)

$0.60

via together

Long (8K tok)

$2.04

via together

Performance Estimates

Throughput by GPU

B200 SXM
280.0 tok/s
B100 SXM
280.0 tok/s
GB200 NVL72 (per GPU)
280.0 tok/s

VRAM Breakdown (B200 SXM, FP8)

Weights
Weights 109.0 GBKV-Cache 1.6 GBActivations 16.0 GBOverhead 5.5 GB

Precision Impact

bf16

218.0 GB

weights/GPU

fp8

109.0 GB

weights/GPU

~280.0 tok/s

int4

54.5 GB

weights/GPU

Quality Benchmarks

Above Average
77th percentile across all models
MMLU
79.0
Average (56th pctile)
HumanEval
55.0
Average (57th pctile)
GSM8K
85.0
Average (55th pctile)
MT-Bench
81.0
Bottom 25% (0th pctile)

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

vllmsglangtensorrt-llm

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy Llama 4 Scout

Similar Models

Command A

111B params · dense

Quality: 81

from $10.00/M

Smaller context, Higher quality, More expensiveCompare →

Command R+

104B params · dense

Quality: 68

from $2.00/M

Smaller context, More expensiveCompare →

Yi-Large

102.6B params · moe

Quality: 74

from $3.00/M

Smaller context, More expensiveCompare →

Inflection 3

100B params · dense

Quality: 74

from $15.00/M

Smaller context, More expensiveCompare →
Smaller context, Lower qualityCompare →

Frequently Asked Questions

How much VRAM does Llama 4 Scout need for inference?

Llama 4 Scout requires approximately 218.0 GB of VRAM at BF16 precision, 109.0 GB at FP8, or 54.5 GB at INT4 quantization. Additional VRAM is needed for KV-cache (196608 bytes per token) and activations (~2.00 GB).

What is the best GPU for Llama 4 Scout?

The top recommended GPU for Llama 4 Scout is the B200 SXM using FP8 precision. It achieves approximately 280.0 tokens/sec at an estimated cost of $4261/month ($5.79/M tokens). Score: 100/100.

How much does Llama 4 Scout inference cost?

Llama 4 Scout API inference starts from $0.18/M input tokens and $0.30/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.