← CanIRunLocal · All hardware · All models

Can I run Llama 3.2 1B locally?

Meta · 1.24B parameters · chat.

Ollama weight files: Q5_K_M 0.91 GB (llama3.2:1b-instruct-q5_K_M); Q8_0 1.3 GB (llama3.2:1b); Q4_K_M 0.81 GB (llama3.2:1b-instruct-q4_K_M).

Yes on 62 machines in this list. The smallest is M1 8GB: Q8_0, about 2 GB, roughly 42 tokens per second, via Ollama. ollama run llama3.2:1b

DGX Spark 128GB: Q8_0, ~2 GB, fits in fast memory, ~158 tok/s — ollama run llama3.2:1b
Mac Studio M5 Max 128GB: Q8_0, ~2 GB, fits in fast memory, ~378 tok/s — ollama run llama3.2:1b
Mac Studio M5 Ultra 256GB: Q8_0, ~2 GB, fits in fast memory, ~738 tok/s — ollama run llama3.2:1b

Setup on DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte)

Ollama: ollama run llama3.2:1b
LM Studio: lms get unsloth/Llama-3.2-1B-Instruct-GGUF@Q8_0
llama.cpp: llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:Q8_0 -c 8192
vLLM: vllm serve meta-llama/Llama-3.2-1B-Instruct --max-model-len 8192 --gpu-memory-utilization 0.9

Open Llama 3.2 1B in the checker

Hardware that fits

Hardware	Runtime	Quant	Memory	Fit	Speed	Command
M1 8GB	Ollama	Q8_0	2 GB	fits in fast memory	~42 tok/s	`ollama run llama3.2:1b`
M2 8GB	Ollama	Q8_0	2 GB	fits in fast memory	~62 tok/s	`ollama run llama3.2:1b`
M3 8GB	Ollama	Q8_0	2 GB	fits in fast memory	~62 tok/s	`ollama run llama3.2:1b`
RTX 3060 Ti 8GB	Ollama	Q8_0	2 GB	fits in fast memory	~207 tok/s	`ollama run llama3.2:1b`
RTX 3070 8GB	Ollama	Q8_0	2 GB	fits in fast memory	~207 tok/s	`ollama run llama3.2:1b`
RTX 4060 8GB	Ollama	Q8_0	2 GB	fits in fast memory	~126 tok/s	`ollama run llama3.2:1b`
RX 7600 8GB	Ollama	Q8_0	2 GB	fits in fast memory	~133 tok/s	`ollama run llama3.2:1b`
RTX 3080 10GB	Ollama	Q8_0	2 GB	fits in fast memory	~351 tok/s	`ollama run llama3.2:1b`
M1 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~42 tok/s	`ollama run llama3.2:1b`
M2 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~62 tok/s	`ollama run llama3.2:1b`
M3 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~62 tok/s	`ollama run llama3.2:1b`
M4 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~74 tok/s	`ollama run llama3.2:1b`
RTX 3060 12GB	Ollama	Q8_0	2 GB	fits in fast memory	~166 tok/s	`ollama run llama3.2:1b`
RTX 4070 12GB	Ollama	Q8_0	2 GB	fits in fast memory	~233 tok/s	`ollama run llama3.2:1b`
RTX 5070 12GB	Ollama	Q8_0	2 GB	fits in fast memory	~310 tok/s	`ollama run llama3.2:1b`
RTX 4060 Ti 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~133 tok/s	`ollama run llama3.2:1b`
RTX 4070 Ti Super 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~310 tok/s	`ollama run llama3.2:1b`
RTX 4080 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~331 tok/s	`ollama run llama3.2:1b`
RTX 5080 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~443 tok/s	`ollama run llama3.2:1b`
RX 7800 XT 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~288 tok/s	`ollama run llama3.2:1b`
RX 9070 XT 16GB	Ollama	Q8_0	2 GB	fits in fast memory	~295 tok/s	`ollama run llama3.2:1b`
M2 24GB	Ollama	Q8_0	2 GB	fits in fast memory	~62 tok/s	`ollama run llama3.2:1b`
M3 24GB	Ollama	Q8_0	2 GB	fits in fast memory	~62 tok/s	`ollama run llama3.2:1b`
M4 24GB	Ollama	Q8_0	2 GB	fits in fast memory	~74 tok/s	`ollama run llama3.2:1b`
RTX 3090 24GB	Ollama	Q8_0	2 GB	fits in fast memory	~432 tok/s	`ollama run llama3.2:1b`
RTX 4090 24GB	Ollama	Q8_0	2 GB	fits in fast memory	~465 tok/s	`ollama run llama3.2:1b`
RX 7900 XTX 24GB	Ollama	Q8_0	2 GB	fits in fast memory	~443 tok/s	`ollama run llama3.2:1b`
M1 Pro 32GB	Ollama	Q8_0	2 GB	fits in fast memory	~123 tok/s	`ollama run llama3.2:1b`
M2 Pro 32GB	Ollama	Q8_0	2 GB	fits in fast memory	~123 tok/s	`ollama run llama3.2:1b`
M4 32GB	Ollama	Q8_0	2 GB	fits in fast memory	~74 tok/s	`ollama run llama3.2:1b`
M3 Pro 36GB	Ollama	Q8_0	2 GB	fits in fast memory	~92 tok/s	`ollama run llama3.2:1b`
Mac Studio M4 Max (32c GPU) 36GB	Ollama	Q8_0	2 GB	fits in fast memory	~252 tok/s	`ollama run llama3.2:1b`
Mac Studio M5 Max (32c GPU) 36GB	Ollama	Q8_0	2 GB	fits in fast memory	~283 tok/s	`ollama run llama3.2:1b`
RTX 5090 32GB	Ollama	Q8_0	2 GB	fits in fast memory	~827 tok/s	`ollama run llama3.2:1b`
M4 Pro 48GB	Ollama	Q8_0	2 GB	fits in fast memory	~168 tok/s	`ollama run llama3.2:1b`
Mac Studio M4 Max (40c GPU) 48GB	Ollama	Q8_0	2 GB	fits in fast memory	~336 tok/s	`ollama run llama3.2:1b`
Mac Studio M5 Max (40c GPU) 48GB	Ollama	Q8_0	2 GB	fits in fast memory	~378 tok/s	`ollama run llama3.2:1b`
M1 Max 64GB	Ollama	Q8_0	2 GB	fits in fast memory	~246 tok/s	`ollama run llama3.2:1b`
M2 Max 64GB	Ollama	Q8_0	2 GB	fits in fast memory	~246 tok/s	`ollama run llama3.2:1b`
M3 Max 64GB	Ollama	Q8_0	2 GB	fits in fast memory	~246 tok/s	`ollama run llama3.2:1b`
M4 Max 64GB	Ollama	Q8_0	2 GB	fits in fast memory	~336 tok/s	`ollama run llama3.2:1b`
Mac Studio M4 Max (40c GPU) 64GB	Ollama	Q8_0	2 GB	fits in fast memory	~336 tok/s	`ollama run llama3.2:1b`
Mac Studio M5 Max (40c GPU) 64GB	Ollama	Q8_0	2 GB	fits in fast memory	~378 tok/s	`ollama run llama3.2:1b`
DGX Spark 64GB (ASUS, Dell, HP, MSI, Acer, Gigabyte)	Ollama	Q8_0	2 GB	fits in fast memory	~158 tok/s	`ollama run llama3.2:1b`
Mac Studio M3 Ultra 96GB	Ollama	Q8_0	2 GB	fits in fast memory	~504 tok/s	`ollama run llama3.2:1b`
Mac Studio M5 Ultra 96GB	Ollama	Q8_0	2 GB	fits in fast memory	~738 tok/s	`ollama run llama3.2:1b`
M3 Max 128GB	Ollama	Q8_0	2 GB	fits in fast memory	~246 tok/s	`ollama run llama3.2:1b`
M4 Max 128GB	Ollama	Q8_0	2 GB	fits in fast memory	~336 tok/s	`ollama run llama3.2:1b`
Mac Studio M4 Max (40c GPU) 128GB	Ollama	Q8_0	2 GB	fits in fast memory	~336 tok/s	`ollama run llama3.2:1b`
Mac Studio M5 Max (40c GPU) 128GB	Ollama	Q8_0	2 GB	fits in fast memory	~378 tok/s	`ollama run llama3.2:1b`
DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte)	Ollama	Q8_0	2 GB	fits in fast memory	~158 tok/s	`ollama run llama3.2:1b`
M2 Ultra 192GB	Ollama	Q8_0	2 GB	fits in fast memory	~492 tok/s	`ollama run llama3.2:1b`
2x Mac Studio M5 Max (40c GPU) 128GB	llama.cpp	Q8_0 GGUF	2 GB	fits in fast memory	~316 tok/s	`llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node`
Mac Studio M3 Ultra 256GB	Ollama	Q8_0	2 GB	fits in fast memory	~504 tok/s	`ollama run llama3.2:1b`
Mac Studio M5 Ultra (64c GPU) 256GB	Ollama	Q8_0	2 GB	fits in fast memory	~738 tok/s	`ollama run llama3.2:1b`
Mac Studio M5 Ultra (80c GPU) 256GB	Ollama	Q8_0	2 GB	fits in fast memory	~738 tok/s	`ollama run llama3.2:1b`
2x DGX Spark 128GB	llama.cpp	Q8_0 GGUF	2 GB	fits in fast memory	~132 tok/s	`llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node`
4x Mac Studio M5 Max (40c GPU) 128GB	llama.cpp	Q8_0 GGUF	2 GB	fits in fast memory	~279 tok/s	`llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node`
2x Mac Studio M5 Ultra (80c GPU) 256GB	llama.cpp	Q8_0 GGUF	2 GB	fits in fast memory	~618 tok/s	`llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node`
Mac Studio M3 Ultra (80c GPU) 512GB	Ollama	Q8_0	2 GB	fits in fast memory	~504 tok/s	`ollama run llama3.2:1b`
4x DGX Spark 128GB	llama.cpp	Q8_0 GGUF	2 GB	fits in fast memory	~116 tok/s	`llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node`
4x Mac Studio M5 Ultra (80c GPU) 256GB	llama.cpp	Q8_0 GGUF	2 GB	fits in fast memory	~545 tok/s	`llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node`

Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-16.

A project by OnDevice Labs · llms.txt