Gemma 4 E4B — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
The 'Efficient 4B' (~8B total params, ~4.5B effective compute) is Gemma 4's sweet spot for 6–8 GB VRAM GPUs. Supports text + image + audio input natively. Scores higher than Gemma 3 12B on MMLU despite using far less effective compute — thanks to improved training data and the new multimodal encoder. Ideal for developers who want image OCR, voice transcription, and reasoning all in one model.
Gemma 4 E4B needs about 6 GB of VRAM at Q4 (QAT) — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
VRAM and speed by quantization
Quoted against NVIDIA RTX 4090 (24 GB). Weights plus overhead only: this model's architecture is unpublished, so no KV cache is added.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 3.4 GB | ~154 tok/s | Fits | |
| Q3_K_M 3.41 bpw | 4.2 GB | ~133 tok/s | Fits | |
| Q4_K_M 4.83 bpw | 5.6 GB | ~107 tok/s | Fits | |
| Q5_K_M 5.67 bpw | 6.5 GB | ~95 tok/s | Fits | |
| Q6_K 6.56 bpw | 7.4 GB | ~86 tok/s | Fits | |
| Q8_0 8.50 bpw | 9.3 GB | ~70 tok/s | Fits | |
| F16 16.00 bpw | 16.8 GB | ~41 tok/s | Fits |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Gemma 4 E4B VRAM calculator →
Get Gemma 4 E4B running
The cheapest catalogued GPU that runs Gemma 4 E4B is the Intel Arc B570 (10 GB).
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
How to run Gemma 4 E4B
Install Ollama, then run:
ollama run gemma4:e4bSpecifications
Verified — Checked against the primary source — the model card or the vendor spec page — and corroborated by a second independent source.
- Parameters
- 8 Billion (~4.5B effective)
- Context window
- 128,000
- Architecture
- Dense Transformer + Multimodal Encoder (image + audio, per-layer embeddings)
- Provider
- Licence
- Apache 2.0
- Practical deployment
- Laptop
- Specified at
- Q4 (QAT)
- System RAM
- 12 GB
- Record updated
- 2026-03-31
Commercial use permitted. No usage restrictions beyond attribution.
Quality and use cases
Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.
| Benchmark | Score | Provenance |
|---|---|---|
| MMLU | 79.8 / 100 % | reported |
| HumanEval | 72.4 / 100 % | reported |
| MATH | 58.3 / 100 % | reported |
Can I run Gemma 4 E4B on my GPU?
- Gemma 4 E4B on AMD Radeon RX 7800 XT
- Gemma 4 E4B on AMD Radeon RX 7900 XT
- Gemma 4 E4B on AMD Radeon RX 7900 XTX
- Gemma 4 E4B on AMD Radeon RX 9060 XT 16GB
- Gemma 4 E4B on AMD Radeon RX 9060 XT 8GB
- Gemma 4 E4B on AMD Radeon RX 9070
- Gemma 4 E4B on AMD Radeon RX 9070 XT
- Gemma 4 E4B on Apple M1
- Gemma 4 E4B on Apple M1 Pro
- Gemma 4 E4B on Apple M2
- Gemma 4 E4B on Apple M2 Pro
- Gemma 4 E4B on Apple M3
- Gemma 4 E4B on Apple M3 Pro
- Gemma 4 E4B on Apple M4
- Gemma 4 E4B on Apple M4 Pro
- Gemma 4 E4B on Apple M5
- Gemma 4 E4B on Intel Arc B570
- Gemma 4 E4B on Intel Arc B580
- Gemma 4 E4B on NVIDIA GeForce RTX 3060 (12GB)
- Gemma 4 E4B on NVIDIA GeForce RTX 3070
- Gemma 4 E4B on NVIDIA GeForce RTX 3070 Ti
- Gemma 4 E4B on NVIDIA GeForce RTX 3080 (10GB)
- Gemma 4 E4B on NVIDIA GeForce RTX 3090
- Gemma 4 E4B on NVIDIA GeForce RTX 4060
- Gemma 4 E4B on NVIDIA GeForce RTX 4060 Ti 16GB
- Gemma 4 E4B on NVIDIA GeForce RTX 4070
- Gemma 4 E4B on NVIDIA GeForce RTX 4070 Super
- Gemma 4 E4B on NVIDIA GeForce RTX 4070 Ti
- Gemma 4 E4B on NVIDIA GeForce RTX 4070 Ti Super
- Gemma 4 E4B on NVIDIA GeForce RTX 4080
- Gemma 4 E4B on NVIDIA GeForce RTX 4080 Super
- Gemma 4 E4B on NVIDIA GeForce RTX 4090
- Gemma 4 E4B on NVIDIA GeForce RTX 5060
- Gemma 4 E4B on NVIDIA GeForce RTX 5060 Ti 16GB
- Gemma 4 E4B on NVIDIA GeForce RTX 5060 Ti 8GB
- Gemma 4 E4B on NVIDIA GeForce RTX 5070
- Gemma 4 E4B on NVIDIA GeForce RTX 5070 Ti
- Gemma 4 E4B on NVIDIA GeForce RTX 5080
- Gemma 4 E4B on NVIDIA GeForce RTX 5090
Other Gemma 4 sizes
Gemma 4 E4B — frequently asked questions
How much VRAM does Gemma 4 E4B need?
About 6 GB at Q4 (QAT) — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Gemma 4 E4B run on an RTX 4090 (24 GB)?
Yes. Gemma 4 E4B needs about 6 GB at Q4 (QAT), inside a 24 GB card, at an estimated 107 tokens/sec.
How do I run Gemma 4 E4B locally?
Install Ollama and run `ollama run gemma4:e4b`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.
What other sizes does Gemma 4 come in?
Gemma 4 E2B (4 GB), Gemma 4 E4B (6 GB), Gemma 4 26B-A4B (16 GB), Gemma 4 31B (20 GB), Gemma 4 12B (Unified) (8 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.