GPU VRAM calculator

Run medgemma-4b-it locally: hardware requirements

medgemma-4b-it needs at least 11.1 GB of GPU memory for the model alone. It is a gated repository, so what each extra person adds is not public, so plan for more.

Google, United States. From its files on Hugging Face, checked 2026-11-05.

Context per person: more context lets the model work with longer conversations and more files, and needs more GPU memory for every person. As a guide: 8K for short chat, 32K for chat over documents, 64K for AI agents, whose instructions, tools and files all count.

80 users means about 16 people working at the same moment. That share is a planning assumption; if you know yours, choose the total that matches it.

Work it out for your team: google/medgemma-4b-it

On Hugging Face

Google, United States · 4.3B parameters · License: other · Gated: accept the license on Hugging Face first

What you need for 16 active users

At least 1 × GeForce RTX 5070

That holds the model alone. What each person adds is not public, so plan for more.

The server

32 GB RAM
32 CPU cores
250 GB NVMe

Rules of thumb

GPU memory in total
at least 11.1 GB
Weights
8.6 GB (published files)
Conversation cache, all active
Not public for a gated model
Engine (vLLM)
3.0 GB

Your options by card

80 users means about 16 using the model at any moment. The others are not waiting; they are simply doing something else. If more become active than a build handles, they queue.

CardMemoryActive users, one serverFor your 16 active
A10080 GBNot public for this modelAt least 1 × A100
B200180 GBNot public for this modelAt least 1 × B200
GeForce RTX 309024 GBNot public for this modelAt least 1 × GeForce RTX 3090
GeForce RTX 409024 GBNot public for this modelAt least 1 × GeForce RTX 4090
GeForce RTX 507012 GBNot public for this modelAt least 1 × GeForce RTX 5070
GeForce RTX 508016 GBNot public for this modelAt least 1 × GeForce RTX 5080
GeForce RTX 509032 GBNot public for this modelAt least 1 × GeForce RTX 5090
H10080 GBNot public for this modelAt least 1 × H100
H200 NVL141 GBNot public for this modelAt least 1 × H200 NVL
L40S48 GBNot public for this modelAt least 1 × L40S
RTX 4000 Ada20 GBNot public for this modelAt least 1 × RTX 4000 Ada
RTX 6000 Ada48 GBNot public for this modelAt least 1 × RTX 6000 Ada
RTX A500024 GBNot public for this modelAt least 1 × RTX A5000
RTX PRO 4000 Blackwell24 GBNot public for this modelAt least 1 × RTX PRO 4000 Blackwell
RTX PRO 4500 Blackwell32 GBNot public for this modelAt least 1 × RTX PRO 4500 Blackwell
RTX PRO 5000 Blackwell48 GBNot public for this modelAt least 1 × RTX PRO 5000 Blackwell
RTX PRO 6000 Blackwell96 GBNot public for this modelAt least 1 × RTX PRO 6000 Blackwell

These are memory limits. A card can hold more people than it can answer quickly, so test your own workload before you buy. More than one card in a server needs an engine that splits the model.

medgemma-4b-it by team size

AI agents, 1 user in 5 active at a time, 64K context, as published.

Total usersAt onceGPU memoryWhat you need
204at least 11.1 GBAt least 1 × GeForce RTX 5070
8016at least 11.1 GBAt least 1 × GeForce RTX 5070
12024at least 11.1 GBAt least 1 × GeForce RTX 5070
500100at least 11.1 GBAt least 1 × GeForce RTX 5070

All models

How this is worked out

GPU memory = weights + a conversation cache for each person + the serving engine, from this model's files on Hugging Face. These are memory limits, not speed. The full method and its limits.

Questions

How much GPU memory does medgemma-4b-it need?

medgemma-4b-it needs at least 11.1 GB of GPU memory for the model alone. It is a gated repository, so what each extra person adds is not public, so plan for more.

Can medgemma-4b-it run on a 24 GB GPU such as the RTX 4090?

Yes as published for the model alone: it needs at least 11.1 GB.

What license does medgemma-4b-it use?

Hugging Face lists its license as other, and it is gated, so you accept the terms there first. Check the license on the model's page before you use it.

Running a model on your own servers?

NeueCode is private autonomous intelligence for enterprise: it takes on your organization's tasks on your own GPUs, with nothing leaving your network, a person approving every action, and all of it on record. It runs each model on one card, so read the single-card figures above.

What you need to run NeueCode