OCR / VisionGemmaMIT

MedGemma 1.5 4B

google/medgemma-1.5-4b-it

Parameters

4.3B

Context

128K

Downloads

315.8K

Likes

776

Architecture

transformer

View on HuggingFace

GPU Requirements

VRAM / RAM / disk estimates by quantization and usage scenario

QuantizationMinimumRecommendedProduction
FP1616bit
13.7 GB
RAM 32.0 · Disk 51.5
16.3 GB
RAM 32.0 · Disk 77.5
32.7 GB
RAM 49.0 · Disk 116.5
FP88bit
8.6 GB
RAM 32.0 · Disk 45.3
11.1 GB
RAM 32.0 · Disk 71.3
27.1 GB
RAM 40.7 · Disk 110.3
INT88bit
8.6 GB
RAM 32.0 · Disk 45.3
11.1 GB
RAM 32.0 · Disk 71.3
27.1 GB
RAM 40.7 · Disk 110.3
AWQ4bit
6.2 GB
RAM 32.0 · Disk 42.2
8.5 GB
RAM 32.0 · Disk 68.2
24.5 GB
RAM 36.7 · Disk 107.2
GPTQ4bit
6.2 GB
RAM 32.0 · Disk 42.3
8.5 GB
RAM 32.0 · Disk 68.3
24.5 GB
RAM 36.8 · Disk 107.3
GGUF4bit
6.3 GB
RAM 32.0 · Disk 42.4
8.6 GB
RAM 32.0 · Disk 68.4
24.6 GB
RAM 36.9 · Disk 107.4

GPU Recommendations

Best ValueBEST

1× RTX 3080 Ti

12 GB VRAM

$79.5/mo

≈ 438.5 tok/s

Effective production throughput ≈ 307 tok/s

score 0.6837

Deploy on vast ↗
Cheapest

1× RTX 3080

10 GB VRAM

$58.6/mo

≈ 365.4 tok/s

Effective production throughput ≈ 255.8 tok/s

score 0.6675

Deploy on vast ↗
Performance

8× RTX 5090

256 GB VRAM

$2,598/mo

≈ 2756.9 tok/s

Effective production throughput ≈ 1929.8 tok/s

score 0.3468

Deploy on vast ↗
Min Complexity

1× RTX 3090

24 GB VRAM

$88.4/mo

≈ 450 tok/s

Effective production throughput ≈ 315 tok/s

score 0.5893

Deploy on vast ↗

💬 Community — real deployment experience

0 Articles · 0 Benchmarks

View Community →

GPUs that fit (single card)

RTX 3080

10 GB VRAM

$58.6

/mo · vast

Tesla V100

16 GB VRAM

$64.8

/mo · vast

RTX 3080 Ti

12 GB VRAM

$79.5

/mo · vast

RTX 3090

24 GB VRAM

$88.4

/mo · vast

RTX 5070

12 GB VRAM

$111

/mo · vast

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