Gemma on NVIDIA Blackwell

Can I run Gemma 3 12B on an RTX 5070?

Yes. Gemma 3 12B at Q4_K_M uses 8.6 GB of the 10.5 GB available on a RTX 5070, and runs at about 58.0 tokens per second. You can push the context to 38K.

Runs comfortably

8.6 GB of 10.5 GB · 82%
010 GB
Weights 7.0 GB
KV cache 0.8 GB
Runtime overhead 0.8 GB

Gemma 3 12B at Q4_K_M leaves 1.9 GB spare on a RTX 5070. There is room to raise the context length or move up a quantisation level.

Generation58.0tok/s
Prompt processing1067tok/s
Max context38Ktokens
KV per 1K tokens0GB

Every quantisation of Gemma 3 12B on a RTX 5070

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 22.7 GB 24.4 GB Won't fit — 2.74
INT8 / W8A8 8.50 11.8 GB 13.5 GB Won't fit — 9.97
Q8_0 (GGUF) 8.50 11.8 GB 13.5 GB Won't fit — 9.97
FP8 (E4M3) 8.00 11.4 GB 13.0 GB Won't fit — 11.7
Q6_K 6.56 9.3 GB 10.9 GB Won't fit 896 28.0
Q5_K_M 5.67 8.1 GB 9.7 GB Fits, but tight 21K 50.8
Q5_K_S 5.52 7.8 GB 9.5 GB Fits, but tight 24K 52.1
AWQ 4-bit 4.25 7.4 GB 9.0 GB Runs comfortably 31K 54.8
GPTQ 4-bit 4.25 7.4 GB 9.0 GB Runs comfortably 31K 54.8
MXFP4 4.25 7.4 GB 9.0 GB Runs comfortably 31K 54.8
Q4_K_M 4.85 7.0 GB 8.6 GB Runs comfortably 38K 58.0
Q4_K_S 4.58 6.6 GB 8.2 GB Runs comfortably 44K 60.8
Q4_0 4.55 6.6 GB 8.2 GB Runs comfortably 44K 61.1
IQ4_XS 4.25 6.2 GB 7.8 GB Runs comfortably 51K 64.6
Q3_K_M 3.91 5.7 GB 7.4 GB Runs comfortably 58K 69.1
IQ3_M 3.70 5.5 GB 7.1 GB Runs comfortably 62K 72.1
IQ3_XXS 3.06 4.6 GB 6.3 GB Runs comfortably 75K 83.5
Q2_K 2.63 4.1 GB 5.7 GB Runs comfortably 84K 93.3
IQ2_XXS 2.06 3.3 GB 5.0 GB Runs comfortably 96K 111
IQ1_M 1.75 2.9 GB 4.6 GB Runs comfortably 103K 123

Also worth checking

Gemma 3 12B on every GPU
Full hardware tableEverything that runs on a RTX 5070
Full model tableWhy the KV cache is the part that bites
Guide