AI Models for Pixel 8 Pro — What runs on 12GB

21 great · 9 slow · 18 won't fit
CHIP
Google Tensor G3
MEMORY BANDWIDTH
68.3 GB/s
NPU
RAM OPTIONS
12 GB
USABLE FOR MODELS
~9 GB
YEAR
2023

Specs checked against manufacturer and public documentation on .

What runs on the Pixel 8 Pro

All 48 models at their recommended quant, on the 12GB configuration. Tap a model for the full report.

ModelParamsQuantNeedsSpeedVerdict
SQwen3 0.6B0.6BQ8_01.3 GB~51.2 tokens/s Runs great
SQwen3 1.7B1.7BQ8_02.6 GB~17.1 tokens/s Runs great
SLlama 3.2 1B1.2BQ4_K_M1.5 GB~38.4 tokens/s Runs great
SLlama 3.2 3B3.2BQ4_K_M2.9 GB~15.4 tokens/s Runs great
SGemma 3 1B1BQ4_K_M1.5 GB~38.4 tokens/s Runs great
SDeepSeek R1 Distill 1.5B1.8BQ4_K_M1.9 GB~27.9 tokens/s Runs great
SSmolLM2 1.7B1.7BQ4_K_M1.9 GB~27.9 tokens/s Runs great
SSmolLM3 3B3.1BQ4_K_M2.8 GB~16.2 tokens/s Runs great
SQwen 3.5 2B2BQ4_K_M2.1 GB~23.6 tokens/s Runs great
STernary Bonsai 4B4BPQ2_02 GB~27.9 tokens/s Runs great
STernary Bonsai 1.7B1.7BPQ2_01.2 GB~61.5 tokens/s Runs great
SLFM2.5 8B-A1B8BQ4_K_M6.6 GB~47.3 tokens/s Runs great
SMinistral 3 3B3BQ4_K_M3 GB~14.6 tokens/s Runs great
STernary Bonsai 8B8BPQ2_03.5 GB~14 tokens/s Runs great
SQwen3 4B4BQ4_K_M3.5 GB~12.3 tokens/s Runs great
SGemma 3 4B4.3BQ4_K_M3.5 GB~12.3 tokens/s Runs great
SPhi-4 Mini 3.8B3.8BQ4_K_M3.5 GB~12.3 tokens/s Runs great
SQwen 3.5 4B4BQ4_K_M3.7 GB~11.4 tokens/s Runs great
SNemotron 3 Nano 4B4BQ4_K_M3.8 GB~11 tokens/s Runs great
SGemma 4 E2B2BQ4_K_M4 GB~9.9 tokens/s Runs great
ABonsai 27B (1-bit)27BQ1_06.5 GB~8.1 tokens/s Runs great
AMistral 7B v0.37.2BQ4_K_M5.7 GB~7 tokens/s! Runs, barely
ALlama 3.1 8B8BQ4_K_M6.3 GB~6.3 tokens/s! Runs, barely
AMinistral 8B8BQ4_K_M6.3 GB~6.3 tokens/s! Runs, barely
ADeepSeek R1 Distill 7B7.6BQ4_K_M6.1 GB~6.5 tokens/s! Runs, barely
AQwen3 8B8.2BQ4_K_M6.4 GB~6.1 tokens/s! Runs, barely
AGemma 4 E4B4BQ4_K_M6.1 GB~6.1 tokens/s! Runs, barely
AMinistral 3 8B8BQ4_K_M6.6 GB~5.9 tokens/s! Runs, barely
AQwen 3.5 9B9BQ4_K_M7.2 GB~5.4 tokens/s! Runs, barely
CGemma 4 12B12BQ4_08.8 GB~4.4 tokens/s! Runs, barely
FGemma 3 12B12.2BQ4_K_M9.1 GB Won't fit
FTernary Bonsai 27B27BPQ2_010.1 GB Won't fit
FMinistral 3 14B14BQ4_K_M10.2 GB Won't fit
FQwen3 14B14.8BQ4_K_M11.1 GB Won't fit
FPhi-4 14B14.7BQ4_K_M11.2 GB Won't fit
FGPT-OSS 20B21BMXFP414.8 GB Won't fit
FGemma 4 26B-A4B26BQ4_017.5 GB Won't fit
FQwen3 30B A3B30.5BQ4_K_M22.3 GB Won't fit
FQwen 3.5 27B27BQ4_K_M20 GB Won't fit
FQwen 3.5 35B-A3B35BQ4_K_M26.2 GB Won't fit
FQwen 3.6 27B27BQ4_K_M20.1 GB Won't fit
FQwen 3.6 35B-A3B35BQ4_K_M26.3 GB Won't fit
FNemotron 3 Nano 30B-A3B30BQ4_K_M28.5 GB Won't fit
FLlama 3.3 70B70BQ4_K_M50.1 GB Won't fit
FQwen3 32B32.8BQ4_K_M23.7 GB Won't fit
FGemma 4 31B31BQ4_K_M22 GB Won't fit
FHunyuan 3 (Hy3)298.8BQ4_K_M212.8 GB Won't fit
FInkling952.4BQ8_0966.9 GB Won't fit

~ = bandwidth-based estimate · ✓ = measured on real hardware

Best model by use case

BEST FOR CHAT

Top everyday assistant & writing pick here — ~51.2 tokens/s at Q8_0, using 1.3 of ~9GB.

BEST FOR CODING

Top code completion & explain-this pick here — ~12.3 tokens/s at Q4_K_M, using 3.5 of ~9GB.

BEST FOR REASONING

Top math & step-by-step thinking pick here — ~27.9 tokens/s at Q4_K_M, using 1.9 of ~9GB.

FAQ

What is the biggest AI model the Pixel 8 Pro can run?

Bonsai 27B (1-bit) (27B parameters) at Q1_0 — it needs 6.5GB of the ~9GB usable on the 12GB Pixel 8 Pro and runs at ~8.1 tokens/s.

How much of the Pixel 8 Pro's 12GB RAM can AI models actually use?

About 9GB. Android keeps roughly 2–4GB for the system and resident apps, so of the 12GB about 9GB is actually available to a model.

Can the Pixel 8 Pro run Llama 3.1 8B?

Yes — at Q4_K_M it needs 6.3GB of the ~9GB usable and runs at ~6.3 tokens/s.

How fast is local AI on the Pixel 8 Pro?

The Google Tensor G3 has 68.3GB/s of memory bandwidth, which is what decode speed scales with. Small models like Ternary Bonsai 1.7B reach ~61.5 tokens/s; larger 7–14B models land in the single digits. Anything above ~8 tokens/s feels smooth for chat.

Which quantization should I use on the Pixel 8 Pro?

Q4_K_M is the size/quality sweet spot for most models. For example, Qwen3 0.6B at Q8_0 takes 1.3GB of memory here. Only drop to Q3 or IQ4 if a model just misses fitting; Q8 rarely pays off on 12GB of RAM.

Is 12GB of RAM enough for local AI?

30 of the 48 models we track fit on the Pixel 8 Pro — 21 run great and 9 run with compromises. 18 models (mostly 12B+) don't fit at their recommended quant.

Other Google phones

Pixel 9 ProPixel 9Pixel 8aPixel 10Pixel 10 ProPixel 9aPixel 9 Pro XLPixel 10 Pro XLPixel 10a