Gemma on NVIDIA Blackwell

Can I run Gemma 3 27B on an RTX 5070 Ti?

Not at Q4_K_M — it needs 17.5 GB against 14.2 GB available. Drop to IQ3_M and it fits, at about 45.3 tokens per second.

Won't fit

17.5 GB of 14.2 GB · 123%
018 GB
Weights 15.6 GB
KV cache 1.1 GB
Runtime overhead 0.8 GB
Over the limit 3.3 GB

Short by 3.3 GB. You can run it with 48 of 62 layers on the RTX 5070 Ti and the rest in system RAM, at roughly 9.17 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation9.17tok/s
Prompt processing674tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Gemma 3 27B on a RTX 5070 Ti

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 51.0 GB 53.0 GB Won't fit 1.03
INT8 / W8A8 8.50 26.8 GB 28.7 GB Won't fit 2.64
Q8_0 (GGUF) 8.50 26.8 GB 28.7 GB Won't fit 2.64
FP8 (E4M3) 8.00 25.5 GB 27.4 GB Won't fit 2.84
Q6_K 6.56 20.9 GB 22.9 GB Won't fit 4.22
Q5_K_M 5.67 18.1 GB 20.0 GB Won't fit 6.03
Q5_K_S 5.52 17.6 GB 19.5 GB Won't fit 6.45
Q4_K_M 4.85 15.6 GB 17.5 GB Won't fit 9.17
AWQ 4-bit 4.25 15.5 GB 17.4 GB Won't fit 9.74
GPTQ 4-bit 4.25 15.5 GB 17.4 GB Won't fit 9.74
MXFP4 4.25 15.5 GB 17.4 GB Won't fit 9.74
Q4_K_S 4.58 14.8 GB 16.7 GB Won't fit 11.5
Q4_0 4.55 14.7 GB 16.6 GB Won't fit 12.3
IQ4_XS 4.25 13.8 GB 15.7 GB Won't fit 16.4
Q3_K_M 3.91 12.7 GB 14.7 GB Won't fit 3K 26.7
IQ3_M 3.70 12.1 GB 14.0 GB Fits, but tight 10K 45.3
IQ3_XXS 3.06 10.2 GB 12.1 GB Runs comfortably 33K 53.2
Q2_K 2.63 8.9 GB 10.8 GB Runs comfortably 48K 60.3
IQ2_XXS 2.06 7.1 GB 9.1 GB Runs comfortably 68K 73.1
IQ1_M 1.75 6.2 GB 8.1 GB Runs comfortably 79K 82.7

Also worth checking