Medgemma 4B IT — Hardware Requirements & GPU Compatibility
Vision
MedGemma 4B is Google's 4.3-billion-parameter instruction-tuned vision-language model, built on Gemma 3 and further trained for medical text and image understanding. It adds a medical image encoder covering chest X-rays, dermatology photos, histopathology slides, and fundus images, and can generate findings and answer questions about medical images. It is a foundation for developers building healthcare applications, not a validated diagnostic tool, and needs further evaluation before clinical use. It runs on a single consumer GPU once quantized. It inherits Gemma 3's 128,000 token context window. It is released under Google's Health AI Developer Foundations terms, a use-restricted license rather than a fully open one. Published in May 2025, it was Google's first instruction-tuned multimodal MedGemma release, alongside a larger text-only 27B variant.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 32K)Best on 24 GB (Q4_K_M · ~231 tok/s · 32K)Macs from 8 GB (Q4_K_M · ~25 tok/s)
Other
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Medgemma 4B IT
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run medgemma:4bQ4_K_M fits fully on the RTX 4090 · 19.0 GB free · 16K context
Download: medgemma-4b-it-Q4_K_M.gguf2.49 GBunsloth
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 24 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on RTX 3060 12GB (7 GB headroom) · Q4_K_M
- Generation speed
- ~82 tok/s
- generation speed
- Cost per 1M output tokens
- $0.20
- per 1M output tokens
How Much VRAM Does Medgemma 4B IT Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 2.0 GB | 6.6 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 32K context · ~326 tok/s | |
| 2.1 GB | 6.7 GB | 3-bit small — 3-bit small quantization | Fits: 32K context · ~317 tok/s | |
| 2.3 GB | 6.9 GB | 3-bit medium — 3-bit medium quantization | Fits: 32K context · ~284 tok/s | |
| 2.4 GB | 7 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 32K context · ~277 tok/s | |
| 2.8 GB | 7.5 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 32K context · ~231 tok/s | |
| 3.4 GB | 8 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 32K context · ~194 tok/s | |
| 3.9 GB | 8.5 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 32K context · ~168 tok/s | |
| 4.7 GB | 9.4 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 32K context · ~139 tok/s |
Selected Q4_K_M
4-bit medium quantization5.0GBat 16K context
- Weights
- 2.2 GB
- KV cache
- 2.5 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 19.0 GB free · ~231 tok/s
≥50 tok/s needs ≥218 GB/s · RTX 5090 ~410, RTX 3090 Ti ~231
Which Hardware Can Run Medgemma 4B IT?
Q4_K_M · 2.8 GBMedgemma 4B IT (Q4_K_M) requires 2.8 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 4+ GB is recommended. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run Medgemma 4B IT, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB).
≤ 11 GB
≤ 11 GB tier →fits · 3.0 GB free · up to 32K ctx
16 GPUs · 1 Mac class · 2 device families
RTX 3080~174Simulate output on RTX 3080RTX 2080 Ti~141Simulate output on RTX 2080 TiRTX 3070 Ti~139Simulate output on RTX 3070 TiM3*8 GB~25Simulate output on iMac 24" M3 (8 GB)Orin Nano~23Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)iPhone~17Simulate output on iPhone 1720 GB
fits · 15.0 GB free · up to 32K ctx
1 GPU
RX 7900 XT~169Simulate output on RX 7900 XT32–40 GB
fits · 27.0 GB free · up to 32K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~410Simulate output on RTX 5090M5 Max36 GB~113Simulate output on Mac Studio M5 Max (36 GB)M532 GB~38Simulate output on MacBook Air 13" M5 (32 GB)Orin~47Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~31Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to Medgemma 4B IT
Gemma 4 family · 35 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Spark X2.6 4BXHToken | — | 4.3 GB | ~221 | same size |
| LightOnOCR 3 4Blightonai | — | 3.4 GB | ~205 | same size |
| Granite Vision 4.1 4BIBM | — | 4 GB | ~228 | same size |
| Qwen3 VL 4B InstructAlibaba | — | 4.5 GB | ~208 | same size |
| MiniCPM V 4OpenBMB | — | 3.1 GB | ~236 | same size |
| Phi 3.5 Vision InstructMicrosoft | — | 9.2 GB | ~183 | same size |
Specifications
- Publisher
- Family
- Gemma 4
- Parameters
- 4.3B
- Release Date
- 2025-05-19
- License
- Other
Frequently Asked Questions
How much VRAM does Medgemma 4B IT need?
Medgemma 4B IT requires 2.8 GB of VRAM at Q4_K_M, or 9.5 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 4.3B × 4.8 bits ÷ 8 = 2.6 GB
KV Cache + Overhead ≈ 0.2 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
What's the best quantization for Medgemma 4B IT?
For Medgemma 4B IT, Q4_K_M (2.8 GB) offers the best balance of quality and VRAM usage. Q4_K_L (2.9 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 1.3 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Medgemma 4B IT on a Mac?
Yes — Mac Mini M2 (8 GB) and 145 other Macs can run Medgemma 4B IT. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 2.8 GB of usable unified memory (RAM minus macOS overhead).
Can I run Medgemma 4B IT locally?
Yes — Medgemma 4B IT can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Medgemma 4B IT?
At Q4_K_M, Medgemma 4B IT can reach ~1690 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~231 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 2.84 × 0.65 = ~1831 tok/s
Estimated speed at Q4_K_M (2.8 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Medgemma 4B IT?
At Q4_K_M, the download is about 2.58 GB. The full-precision BF16 version is 8.60 GB. The smallest option (IQ2_XXS) is 1.18 GB.
Which GPUs can run Medgemma 4B IT?
52 consumer GPUs can run Medgemma 4B IT at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run Medgemma 4B IT?
163 devices with unified memory can run Medgemma 4B IT at Q4_K_M (2.8 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.