← CanIRunLocal · All hardware · All models
Yes, with partial offload (slower). Gemma 4 26B-A4B (MoE) on RTX 3080 12GB at Q4_K_M uses system RAM (19.6 GB needed, 11.5 GB usable), about 17 tok/s (estimate).
Hardware: RTX 3080 12GB · Model: Gemma 4 26B-A4B (MoE)
RTX 3080 12GB has 12 GB of VRAM and 912 GB/s of memory bandwidth. The checker holds back 0.5 GB of VRAM for the driver and assumes 16 GB of system RAM, so fast memory usable by Ollama is 11.5 GB. Partial offload may also use 60% of that system RAM. Gemma 4 26B-A4B (MoE) (25.8B, 3.8B active, Google) at Q4_K_M is a 18 GB download and needs 19.6 GB with headroom for about 8k tokens.
| Best quant | Q4_K_M |
|---|---|
| Download size | 18 GB |
| Memory needed | 19.6 GB |
| Usable fast memory | 11.5 GB |
| Fit | partial offload (slower) |
| Estimated speed | ~17 tok/s (estimate) |
| Command | ollama run gemma4:26b-a4b-it-q4_K_M |
Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-23.
A project by OnDevice Labs · llms.txt