Autor: Jakub Rusinowski · Ostatnia aktualizacja: 31 marca 2026
Model library → Gemma 4 → Gemma 4 26B-A4B
Gemma 4's Mixture-of-Experts variant: 26B total parameters but only 4B active per forward pass. This gives 12B-class quality at 4B-class inference speed. Supports text, image, and audio. Needs ~16–18 GB at Q4_K_M — fits in an RTX 3090 or 4090. Excellent for developers who need a step above the E4B without occupying the full GPU.
Gemma 4 26B-A4B needs about 16 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 | 26 Billion (4B active per token) |
| Context window | 256,000 |
| Architecture | Mixture-of-Experts + Multimodal Encoder (image + audio) |
| Provider | |
| Licence | Apache 2.0 |
| Specified at | Q4_K_M |
| System RAM | 32 GB |
| Record updated | 2026-03-31 |
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 | 8.5 GB | 9.3 GB | ~195 tok/s (est.) | Fits comfortably |
| Q3_K_M | 11.1 GB | 11.9 GB | ~178 tok/s (est.) | Fits comfortably |
| Q4_K_M | 15.7 GB | 16.5 GB | ~152 tok/s (est.) | Fits comfortably |
| Q5_K_M | 18.4 GB | 19.2 GB | ~140 tok/s (est.) | Fits comfortably |
| Q6_K | 21.3 GB | 22.1 GB | ~130 tok/s (est.) | Tight fit |
| Q8_0 | 27.6 GB | 28.4 GB | ~17 tok/s (est.) | Offloads to system RAM (slow) |
| F16 | 52.0 GB | 52.8 GB | ~11 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 26B-A4B 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 26B-A4B is the AMD Radeon RX 9060 XT 16GB (16 GB).
Install Ollama, then run:
ollama run gemma4:26b-a4bWeights on Hugging Face: google/gemma-4-26B-A4B-it.
Quality scores as published by the model's authors or an independent evaluator — not throughput, and not measured by us.
| Benchmark | Score | Provenance |
|---|---|---|
| MMLU | 87.5 / 100 % | reported |
| HumanEval | 83.7 / 100 % | reported |
| MATH | 74.2 / 100 % | reported |
Best for: reasoning, multimodal, coding, long context, speed efficient.
← All Gemma 4 models | VRAM calculator | Build a PC for this model | Check your own hardware