OCR / VisionMistralMIT

Mistral Small 4

mistralai/Mistral-Small-4-119B-2603

Parameters

119B

Context

256K

Downloads

136.6K

Likes

414

Architecture

moe_transformer

View on HuggingFace

GPU Requirements

VRAM / RAM / disk estimates by quantization and usage scenario

QuantizationMinimumRecommendedProduction
FP1616bit
284.0 GB
RAM 426.0 · Disk 384.8
298.9 GB
RAM 448.3 · Disk 410.8
327.5 GB
RAM 491.2 · Disk 449.8
FP88bit
143.8 GB
RAM 215.7 · Disk 211.9
152.3 GB
RAM 228.5 · Disk 237.9
174.5 GB
RAM 261.8 · Disk 276.9
INT88bit
143.8 GB
RAM 215.7 · Disk 211.9
152.3 GB
RAM 228.5 · Disk 237.9
174.5 GB
RAM 261.8 · Disk 276.9
AWQ4bit
76.3 GB
RAM 114.5 · Disk 128.7
81.8 GB
RAM 122.7 · Disk 154.7
100.9 GB
RAM 151.4 · Disk 193.7
GPTQ4bit
77.2 GB
RAM 115.8 · Disk 129.8
82.7 GB
RAM 124.1 · Disk 155.8
101.9 GB
RAM 152.8 · Disk 194.8
GGUF4bit
78.9 GB
RAM 118.4 · Disk 131.9
84.5 GB
RAM 126.8 · Disk 157.9
103.8 GB
RAM 155.7 · Disk 196.9

GPU Recommendations

Best ValueBEST

1× H200 SXM

141 GB VRAM

$2,555/mo

≈ 83.5 tok/s

Effective production throughput ≈ 58.4 tok/s

score 0.6257

Cheapest

8× Tesla V100

128 GB VRAM

$162/mo

≈ 0 tok/s (N/A — benchmark unavailable)

Effective production throughput ≈ 0 tok/s

score 0.527

Deploy on vast ↗
Performance

8× H200 SXM

1128 GB VRAM

$20,440/mo

≈ 267.2 tok/s

Effective production throughput ≈ 187 tok/s

score 0.2197

Min Complexity

2× RTX A6000

96 GB VRAM

$420/mo

≈ 21.4 tok/s

Effective production throughput ≈ 15 tok/s

score 0.6204

Deploy on vast ↗

💬 Community — real deployment experience

0 Articles · 0 Benchmarks

View Community →

GPUs that fit (single card)

H200 SXM

141 GB VRAM

$2,555

/mo · lambda

B200

180 GB VRAM

$3,013

/mo · vast

B200 SXM

180 GB VRAM

AmciHub

AI Model Deployment & Compute Intelligence — analyze AI model requirements, GPU performance, cloud pricing and deployment costs to find the right deployment solution.

© 2026 AmciHub. All rights reserved. Model → Requirement → GPU → Benchmark → Cloud → Cost → Recommendation