Skip to content
Updated minutes ago
Mistral

Mistral Large 2

Mistral AI · dense · 123B parameters · 131,072 context

Quality
75.0

Parameters

123B

Context Window

128K tokens

Architecture

Dense

Best GPU

B200 SXM

Cheapest API

$2.50/M

Quality Score

75/100

Intelligence Brief

Mistral Large 2 is a 123B parameter DENSE model from Mistral AI, featuring Grouped Query Attention (GQA) with 88 layers and 12,288 hidden dimensions. With a 131,072 token context window, it supports tools, structured output, code, math, multilingual. On standardized benchmarks, it achieves MMLU 84, HumanEval 53, GSM8K 91.2. The most cost-effective API deployment is via together at $2.50/M output tokens. For self-hosted inference, B200 SXM delivers optimal throughput at $4261/month.

Architecture Details

TypeDENSE
Total Parameters123B
Active Parameters123B
Layers88
Hidden Dimension12,288
Attention Heads96
KV Heads8
Head Dimension128
Vocab Size32,768

Memory Requirements

BF16 Weights

246.0 GB

FP8 Weights

123.0 GB

INT4 Weights

61.5 GB

KV-Cache per Token360448 bytes
Activation Estimate3.50 GB

GPU Compatibility Matrix

Mistral Large 2 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

$2.50/M

output tokens

Self-Hosted

Single GPU

B200 SXM

$4261/mo

Min VRAM: 123 GB

Scale

Multi-GPU

H20 x2

280.0 tok/s

TP· $1879/mo

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
together$2.50$2.50
Cheapest
mistral$2.00$6.00
Low Input

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
togetherBest Value$2.50$2.50$25
mistral$2.00$6.00$40

Cost per 1,000 Requests

Short (500 tok)

$1.75

via together

Medium (2K tok)

$7.00

via together

Long (8K tok)

$25.00

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
Act
Weights 123.0 GBKV-Cache 3.0 GBActivations 28.0 GBOverhead 6.2 GB

Precision Impact

bf16

246.0 GB

weights/GPU

fp8

123.0 GB

weights/GPU

~280.0 tok/s

int4

61.5 GB

weights/GPU

Quality Benchmarks

Above Average
80th percentile across all models
MMLU
84.0
Average (72th pctile)
HumanEval
53.0
Average (54th pctile)
GSM8K
91.2
Average (74th pctile)
MT-Bench
84.0
Bottom 25% (0th pctile)

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

vllmsglangtgitensorrt-llm

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy Mistral Large 2

Similar Models

More expensiveCompare →
Higher qualityCompare →

DBRX Base

132B params · moe

Quality: 50

from $2.25/M

Smaller context, Lower qualityCompare →

DBRX Instruct

132B params · moe

Quality: 50

from $1.20/M

Smaller context, Lower quality, CheaperCompare →

Command A

111B params · dense

Quality: 81

from $10.00/M

Larger context, Higher quality, More expensiveCompare →

Frequently Asked Questions

How much VRAM does Mistral Large 2 need for inference?

Mistral Large 2 requires approximately 246.0 GB of VRAM at BF16 precision, 123.0 GB at FP8, or 61.5 GB at INT4 quantization. Additional VRAM is needed for KV-cache (360448 bytes per token) and activations (~3.50 GB).

What is the best GPU for Mistral Large 2?

The top recommended GPU for Mistral Large 2 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 Mistral Large 2 inference cost?

Mistral Large 2 API inference starts from $2.50/M input tokens and $2.50/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.