Gemma 4 12B (Unified) — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
Released June 3, 2026 as a separate follow-up to the March launch, the 12B 'Unified' model uses a novel encoder-free architecture that feeds image, audio, and video directly into the LLM backbone — no bolt-on vision/audio encoders. Google's headline claim is that it runs entirely on a typical 16 GB laptop, making it the most capable Gemma 4 tier most people can actually run. Apache 2.0 licensed.
Gemma 4 12B (Unified) needs about 8 GB of VRAM at Q4_K_M — 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). Includes the KV cache at 8K context, so it reads higher than the headline figure.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 5.4 GB | ~122 tok/s | Fits | |
| Q3_K_M 3.41 bpw | 6.6 GB | ~102 tok/s | Fits | |
| Q4_K_M 4.83 bpw | 8.7 GB | ~79 tok/s | Fits | |
| Q5_K_M 5.67 bpw | 9.9 GB | ~70 tok/s | Fits | |
| Q6_K 6.56 bpw | 11.3 GB | ~62 tok/s | Fits | |
| Q8_0 8.50 bpw | 14.2 GB | ~50 tok/s | Fits | |
| F16 16.00 bpw | 25.4 GB | ~4 tok/s | Offload |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Gemma 4 12B (Unified) VRAM calculator →
Get Gemma 4 12B (Unified) running
The cheapest catalogued GPU that runs Gemma 4 12B (Unified) 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 12B (Unified)
Install Ollama, then run:
ollama run gemma4:12bSpecifications
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
- Parameters
- 12 Billion
- Context window
- 128,000
- Architecture
- Encoder-free Unified Multimodal Transformer (text + image + audio + video)
- Provider
- Licence
- Apache 2.0
- Practical deployment
- Laptop
- Specified at
- Q4_K_M
- System RAM
- 32 GB
- Record updated
- 2026-06-03
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.
Can I run Gemma 4 12B (Unified) on my GPU?
- Gemma 4 12B (Unified) on AMD Radeon RX 7800 XT
- Gemma 4 12B (Unified) on AMD Radeon RX 7900 XT
- Gemma 4 12B (Unified) on AMD Radeon RX 7900 XTX
- Gemma 4 12B (Unified) on AMD Radeon RX 9060 XT 16GB
- Gemma 4 12B (Unified) on AMD Radeon RX 9060 XT 8GB
- Gemma 4 12B (Unified) on AMD Radeon RX 9070
- Gemma 4 12B (Unified) on AMD Radeon RX 9070 XT
- Gemma 4 12B (Unified) on Apple M1
- Gemma 4 12B (Unified) on Apple M1 Pro
- Gemma 4 12B (Unified) on Apple M2
- Gemma 4 12B (Unified) on Apple M2 Pro
- Gemma 4 12B (Unified) on Apple M3
- Gemma 4 12B (Unified) on Apple M3 Pro
- Gemma 4 12B (Unified) on Apple M4
- Gemma 4 12B (Unified) on Apple M4 Pro
- Gemma 4 12B (Unified) on Apple M5
- Gemma 4 12B (Unified) on Intel Arc B570
- Gemma 4 12B (Unified) on Intel Arc B580
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 3060 (12GB)
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 3070
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 3070 Ti
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 3080 (10GB)
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 3090
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4060
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4060 Ti 16GB
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4070
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4070 Super
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4070 Ti
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4070 Ti Super
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4080
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4080 Super
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 4090
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 5060
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 5060 Ti 16GB
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 5060 Ti 8GB
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 5070
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 5070 Ti
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 5080
- Gemma 4 12B (Unified) on NVIDIA GeForce RTX 5090
Other Gemma 4 sizes
Gemma 4 12B (Unified) — frequently asked questions
How much VRAM does Gemma 4 12B (Unified) need?
About 8 GB at Q4_K_M — 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 12B (Unified) run on an RTX 4090 (24 GB)?
Yes. Gemma 4 12B (Unified) needs about 8 GB at Q4_K_M, inside a 24 GB card, at an estimated 79 tokens/sec.
How do I run Gemma 4 12B (Unified) locally?
Install Ollama and run `ollama run gemma4:12b`. 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.