Q3_K_M
14.5 GBMin VRAM: 16.5 GB
Recommended VRAM: 20 GB
Min RAM: 20 GB
Context: 8K / 256K
Loading model details...
Fetching variants, compatibility details, and metadata.
Loading model details...
Fetching variants, compatibility details, and metadata.
Share Gemma 4 31B with someone who is deciding what to run locally.
Social proof
33% of 800 scanned PCs run Gemma 4 31B fully on GPU.
490 keep at least some work on GPU. Based on anonymous compatibility checks.
General-purpose local model brief
Best for
Consider alternatives if
Quantization tip: Benchmark at least two quantizations and validate with a task-specific eval set before production use.
New to local models? Smaller quantization variants are easier to run, while larger ones can improve quality at the cost of more memory.
Q3_K_M
14.5 GBMin VRAM: 16.5 GB
Recommended VRAM: 20 GB
Min RAM: 20 GB
Context: 8K / 256K
Q4_K_M
18.4 GBMin VRAM: 20.5 GB
Recommended VRAM: 24 GB
Min RAM: 24 GB
Context: 8K / 256K
Q8_0
33.2 GBMin VRAM: 35 GB
Recommended VRAM: 40 GB
Min RAM: 40 GB
Context: 8K / 256K
| Quantization | File Size | Min VRAM | Recommended VRAM | Min RAM | Context |
|---|---|---|---|---|---|
| Q3_K_M | 14.5 GB | 16.5 GB | 20 GB | 20 GB | 8K / 256K |
| Q4_K_M | 18.4 GB | 20.5 GB | 24 GB | 24 GB | 8K / 256K |
| Q8_0 | 33.2 GB | 35 GB | 40 GB | 40 GB | 8K / 256K |
These GPUs meet the recommended 20 GB VRAM for the Q3_K_M quantization. Estimated speeds are approximate and assume full GPU offloading.
Budget Pick
AMD Radeon RX 7900 XT20 GB VRAM · ~38 tok/s
Lowest cost that meets recommended VRAM
Check price on AmazonFastest Pick
NVIDIA GeForce RTX 509032 GB VRAM · ~85.2 tok/s
Highest estimated throughput
Check price on AmazonBest Value
NVIDIA GeForce RTX 409024 GB VRAM · ~47.9 tok/s
Best speed per dollar of VRAM
Rent on RunPodNeed a detailed comparison? See all GPU rankings for Gemma 4 31B.