Autor: Jakub Rusinowski · Ostatnia aktualizacja: 3 czerwca 2026
Model library → Gemma 4 → Gemma 4 12B (Unified)
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.
| Parameters | 12 Billion |
| Context window | 128,000 |
| Architecture | Encoder-free Unified Multimodal Transformer (text + image + audio + video) |
| Provider | |
| Licence | Apache 2.0 |
| Specified at | Q4_K_M |
| System RAM | 32 GB |
| Record updated | 2026-06-03 |
Apache-2.0 — commercial use permitted. Commercial use permitted. No usage restrictions beyond attribution.
Modelled on a reference NVIDIA RTX 4090 (24 GB), with no KV cache (this record has no published architecture). Speed figures are ESTIMATES from the memory-bandwidth roofline described on the methodology page, not benchmarks we ran — rows marked measured come from published or reader-submitted runs. VRAM here includes the KV cache, so it reads higher than the headline figure above, which does not.
| Quant | Weights | VRAM needed | Est. speed | Fit on 24 GB |
|---|---|---|---|---|
| Q2_K | 3.9 GB | 4.7 GB | ~120 tok/s (est.) | Fits comfortably |
| Q3_K_M | 5.1 GB | 5.9 GB | ~101 tok/s (est.) | Fits comfortably |
| Q4_K_M | 7.2 GB | 8.0 GB | ~79 tok/s (est.) | Fits comfortably |
| Q5_K_M | 8.5 GB | 9.3 GB | ~70 tok/s (est.) | Fits comfortably |
| Q6_K | 9.8 GB | 10.6 GB | ~62 tok/s (est.) | Fits comfortably |
| Q8_0 | 12.8 GB | 13.6 GB | ~50 tok/s (est.) | Fits comfortably |
| F16 | 24.0 GB | 24.8 GB | ~4 tok/s (est.) | Offloads to system RAM (slow) |
Want the memory numbers alone, at every quantization level and your own context length? Use the Gemma 4 12B (Unified) VRAM calculator.
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.
The cheapest catalogued GPU that runs Gemma 4 12B (Unified) is the Intel Arc B570 (10 GB).
Install Ollama, then run:
ollama run gemma4:12bWeights on Hugging Face: google/gemma-4-12B-it.
Best for: multimodal, video analysis, audio transcription, laptop, coding.
← All Gemma 4 models | VRAM calculator | Check your own hardware