No — not on this device
Gemma 4 26B-A4B at Q4_K_M needs 16.0 GB against 10.7 GB usable, and the shortfall of 5.3 GB is more than 128 GB of system RAM can cover at a tolerable speed. A smaller sibling or a lower quantisation is the honest answer here.
Mixture of experts with 3.8B active. Slower to think than Qwen3.6 35B-A3B, faster to answer, and it sees images.
What hardware do I need for Gemma 4 26B-A4B? →
Fits instead: Gemma 4 12B (8.2 GB) · Gemma 4 E4B (5.2 GB)
The VRAM budget
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 26.2 GB | 27.3 GB | — | ~4.7 | −0.1% ppl | 16.6 GB over |
| Q6_K | 20.2 GB | 21.3 GB | — | ~6.1 | −0.4% ppl | 10.6 GB over |
| Q5_K_M | 17.5 GB | 18.6 GB | — | ~7.1 | −0.8% ppl | 7.9 GB over |
| Q4_K_M | 14.9 GB | 16.0 GB | — | ~8.3 | −1.9% ppl | 5.3 GB over |
| Q3_K_M | 12.1 GB | 13.2 GB | — | ~10 | −5.4% ppl | 2.5 GB over |
| Q2_K | 10.3 GB | 11.4 GB | — | ~12 | −15% ppl | 0.7 GB over |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model interleaves sliding-window layers (1024 tokens, 5 of 30 layers global), which is why its cache barely grows with context.
How to run it
$ pip install mlx-lm $ mlx_lm.generate --model mlx-community/gemma-4-26B-A4B-it-4bit \ --max-tokens 512 --prompt "Hello"
Apple's own array framework. The fastest path on Apple Silicon. More on MLX.