Gemma 4 31B IT at Q4_K_M needs about 20.5 GB, more than the 12 GB of the RTX 4070: part of the model runs from system RAM, which slows generation. At Q3_K_M it does not fit in VRAM.
Q4_K_M · 8K context
The model needs about 20.5 GB but only 11.6 GB of VRAM is usable: 8.8 GB must be loaded into system RAM. It runs, 57% on the GPU, but more slowly.
Estimated speed is at most 4 tokens/s: usable, but slow.
~20.5 GB (19.7 GB–21.3 GB) needed · estimate
8.8 GB must be offloaded to system RAM
EstimateData updated 40 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card
What makes this estimate less reliable
| Gemma 4 31B IT Q8_08K~34.5 GB · 0.9–1.7 tok/s estimate | Not recommended |
|---|---|
| Gemma 4 31B IT Q6_K8K~27.2 GB · 1.3–2.5 tok/s estimate | Possible but slow |
| Gemma 4 31B IT Q5_K_M8K~23.7 GB · 1.6–3 tok/s estimate | Possible but slow |
| Gemma 4 31B IT Q4_K_M8K~20.5 GB · 2–4 tok/s estimate | Possible but slow |
| Gemma 4 31B IT IQ4_XS8K~18.6 GB · 3–5.5 tok/s estimate | CPU/GPU offload |
| Gemma 4 31B IT Q3_K_M8K~17 GB · 3.5–6.5 tok/s estimate | CPU/GPU offload |
Prices and links follow the market you choose.
AMD
CompatibleRuns Gemma 4 31B IT Q4_K_M entirely in memory
Unified memory PC
CompatibleRuns Gemma 4 31B IT Q4_K_M entirely in memory
Lowest memory tier
Compatible with limitsRuns Gemma 4 31B IT Q4_K_M entirely in memory
No current offer imported for this market: the link opens a marketplace search.
