Calculator GPUs Engines Guides

llamaperf/Gemma 3

Gemma 3

Google · 0 reports

Gemma 3 VRAM requirements by size and quant →

No reports yet for Gemma 3.

Recommended hardware

Best GPUs for 30B local LLMs

Best GPUs for running 30B-class local LLMs (Qwen 2.5 32B, Mistral Small) — community-reported tokens-per-second and VRAM fit.

Browse

All GPUs
All models
All engines
Calculator
Compare GPUs
Leaderboard
Best
Best for coding
Macs
Guides
Local model performance
Discussions

Best of

GPUs for running local LLMs
Mac for running local LLMs
NVIDIA GPUs for local LLMs
GPUs for 70B local LLMs
GPUs for 30B local LLMs
Mac for 70B local LLMs

Popular GPUs

RTX 3090
RTX 5090
AMD Strix Halo 128GB
RTX Pro 6000 Blackwell
RTX 5060 Ti 16GB
DGX Spark
RTX 3060 12GB
M5 Max 128GB

Popular models

Qwen3.8
Qwen3.6
DeepSeek V4 Flash
Gemma 4
Muse
GLM-5.3
Qwen3.5
DeepSeek V4

Engines

llama.cpp
vLLM
MLX
Ollama
LM Studio
ExLlamaV3

llamaperf · community-sourced local model performance reports

Submit a report Contact