Autor: Jakub Rusinowski · Ostatnia aktualizacja: 3 kwietnia 2026
Przełomowy model o otwartych wagach od Google DeepMind. Gemma 4 27B zapewnia wydajność na poziomie GPT-4 przy 14 GB VRAM, udostępniając AI klasy frontier na konsumenckich GPU. Wyróżnia się hybrydową architekturą z przeplatanymi warstwami uwagi lokalnej i globalnej, rozumieniem wielu obrazów naraz i oknem kontekstu 128k. Osiąga 85 tokenów na sekundę na RTX 4090.
| Licence | What it permits | Applies to |
|---|---|---|
Gemma Terms | Commercial use permitted Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms. | Gemma 4 4B, Gemma 4 12B, Gemma 4 27B ⭐ |
| Gemma 4 4B | Min 3 GB VRAM · Q4_K_M · 128,000 ctx · ollama run gemma4:4b |
| Gemma 4 12B | Min 8 GB VRAM · Q4_K_M · 128,000 ctx · |
| Gemma 4 27B ⭐ | Min 17 GB VRAM · Q4_K_M · 128,000 ctx · ollama run gemma4:27b |
The cheapest GPU that runs Gemma 4 (Legacy Listing — Unverified) locally (min 3 GB VRAM) is the Intel Arc B570 (10 GB).
Install Ollama then run: ollama run gemma4:4b
Minimum VRAM: 3 GB. For best results use Q4_K_M quantization.
Gemma 4 (Legacy Listing — Unverified) needs about 3 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Gemma 4 4B (3 GB, Q4_K_M); Gemma 4 12B (8 GB, Q4_K_M); Gemma 4 27B ⭐ (17 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Yes — Gemma 4 (Legacy Listing — Unverified) runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
Q4_K_M is the best balance of quality and VRAM for Gemma 4 (Legacy Listing — Unverified) in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run gemma4:4b. This downloads Gemma 4 (Legacy Listing — Unverified) and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.