Best GPU for running LLMs locally

Pick a budget. Each guide ranks the cards we track by how many local AI models they run well, then by speed, using the same estimates as the calculator.

By budget

Under $400Top pick: Intel Arc A770 16GB16 GB · 28/69 models Under $600Top pick: Radeon RX 907016 GB · 28/69 models Under $800Top pick: RTX 5070 Ti16 GB · 28/69 models Under $1,000Top pick: Radeon RX 7900 XTX24 GB · 48/69 models Under $1,600Top pick: Radeon AI PRO R970032 GB · 49/69 models Under $2,000Top pick: RTX 509032 GB · 49/69 models

Head to head

Prefer a Mac?

Apple Silicon shares memory between CPU and GPU, so big-memory Macs run models no single consumer GPU can.

Mac mini M4 (16 GB)

Runs 21/69 models well, up to Mistral Nemo 12B.

Check Mac mini M4 (16 GB) prices Every model

Mac mini M4 Pro (24 GB)

Runs 28/69 models well, up to gpt-oss 20B.

Check Mac mini M4 Pro (24 GB) prices Every model

Mac mini M4 Pro (48 GB)

Runs 49/69 models well, up to Mixtral 8x7B.

Check Mac mini M4 Pro (48 GB) prices Every model

Mac Studio M4 Max (64 GB)

Runs 53/69 models well, up to Qwen2.5 72B.

Check Mac Studio M4 Max (64 GB) prices Every model

MacBook Pro M4 Max (128 GB)

Runs 57/69 models well, up to gpt-oss 120B.

Check MacBook Pro M4 Max (128 GB) prices Every model

Mac Studio M3 Ultra (256 GB)

Runs 60/69 models well, up to DeepSeek V4 Flash 284B-A13B.

Check Mac Studio M3 Ultra (256 GB) prices Every model

Every card we track

GPU	VRAM	Launch price	Runs well	Largest model	Llama 3.1 8B	Qwen3 14B	Qwen3 32B
RTX PRO 6000 Blackwell	96 GB	$8,565	57/69	gpt-oss 120B	~67 tok/s	~36 tok/s	~16 tok/s	Prices
RTX 5090	32 GB	$1,999	49/69	Mixtral 8x7B	~67 tok/s	~36 tok/s	~40 tok/s	Prices
Radeon AI PRO R9700	32 GB	$1,299	49/69	Mixtral 8x7B	~24 tok/s	~13 tok/s	~14 tok/s	Prices
RTX 4090	24 GB	$1,599	48/69	Qwen3.6 35B-A3B	~38 tok/s	~38 tok/s	~30 tok/s	Prices
RTX 3090 Ti	24 GB	$1,999	48/69	Qwen3.6 35B-A3B	~38 tok/s	~38 tok/s	~30 tok/s	Prices
Radeon RX 7900 XTX	24 GB	$999	48/69	Qwen3.6 35B-A3B	~36 tok/s	~37 tok/s	~29 tok/s	Prices
RTX 3090	24 GB	$1,499	48/69	Qwen3.6 35B-A3B	~35 tok/s	~36 tok/s	~28 tok/s	Prices
Radeon RX 7900 XT	20 GB	$899	41/69	Nemotron 3 Nano 30B-A3B	~30 tok/s	~31 tok/s	tight	Prices
RTX 5080	16 GB	$999	28/69	gpt-oss 20B	~68 tok/s	~47 tok/s	tight	Prices
RTX 5070 Ti	16 GB	$749	28/69	gpt-oss 20B	~63 tok/s	~44 tok/s	tight	Prices
RTX 4080 Super	16 GB	$999	28/69	gpt-oss 20B	~52 tok/s	~36 tok/s	tight	Prices
RTX 4080	16 GB	$1,199	28/69	gpt-oss 20B	~50 tok/s	~35 tok/s	tight	Prices
RTX 4070 Ti Super	16 GB	$799	28/69	gpt-oss 20B	~47 tok/s	~33 tok/s	tight	Prices
Radeon RX 9070	16 GB	$549	28/69	gpt-oss 20B	~45 tok/s	~32 tok/s	tight	Prices
Radeon RX 9070 XT	16 GB	$599	28/69	gpt-oss 20B	~45 tok/s	~32 tok/s	tight	Prices
Radeon RX 7800 XT	16 GB	$499	28/69	gpt-oss 20B	~44 tok/s	~31 tok/s	tight	Prices
Intel Arc A770 16GB	16 GB	$349	28/69	gpt-oss 20B	~39 tok/s	~28 tok/s	tight	Prices
Radeon RX 6800 XT	16 GB	$649	28/69	gpt-oss 20B	~36 tok/s	~25 tok/s	tight	Prices
RTX 5060 Ti 16GB	16 GB	$429	28/69	gpt-oss 20B	~32 tok/s	~22 tok/s	tight	Prices
Radeon RX 9060 XT 16GB	16 GB	$349	28/69	gpt-oss 20B	~23 tok/s	~16 tok/s	tight	Prices
Radeon RX 7600 XT	16 GB	$329	28/69	gpt-oss 20B	~20 tok/s	~14 tok/s	tight	Prices
RTX 4060 Ti 16GB	16 GB	$499	28/69	gpt-oss 20B	~20 tok/s	~14 tok/s	tight	Prices
RTX 3080 Ti	12 GB	$1,199	27/69	Qwen3 14B	~64 tok/s	~61 tok/s	too big	Prices
RTX 5070	12 GB	$549	27/69	Qwen3 14B	~47 tok/s	~45 tok/s	too big	Prices
RTX 4070	12 GB	$599	27/69	Qwen3 14B	~35 tok/s	~34 tok/s	too big	Prices
RTX 4070 Super	12 GB	$599	27/69	Qwen3 14B	~35 tok/s	~34 tok/s	too big	Prices
RTX 4070 Ti	12 GB	$799	27/69	Qwen3 14B	~35 tok/s	~34 tok/s	too big	Prices
Intel Arc B580	12 GB	$249	27/69	Qwen3 14B	~32 tok/s	~30 tok/s	too big	Prices
RTX 3060 12GB	12 GB	$329	27/69	Qwen3 14B	~25 tok/s	~24 tok/s	too big	Prices
RTX 2080 Ti	11 GB	$999	22/69	Qwen3 14B	~43 tok/s	~41 tok/s	too big	Prices
RTX 3080 10GB	10 GB	$699	21/69	Mistral Nemo 12B	~69 tok/s	tight	too big	Prices
RTX 5060	8 GB	$299	17/69	Qwen3 8B	~47 tok/s	tight	too big	Prices
RTX 5060 Ti 8GB	8 GB	$379	17/69	Qwen3 8B	~47 tok/s	tight	too big	Prices
RTX 3060 Ti	8 GB	$399	17/69	Qwen3 8B	~47 tok/s	tight	too big	Prices
RTX 3070	8 GB	$499	17/69	Qwen3 8B	~47 tok/s	tight	too big	Prices
RTX 4060 Ti 8GB	8 GB	$399	17/69	Qwen3 8B	~30 tok/s	tight	too big	Prices
RTX 4060	8 GB	$299	17/69	Qwen3 8B	~29 tok/s	tight	too big	Prices

Prices are US launch prices (MSRP) for comparison; street prices change daily. Links go to Amazon, where we earn a commission at no cost to you.

Frequently asked questions

How much VRAM do I need for local LLMs?

8 GB runs 7 to 8B models, 12 GB runs 12 to 14B, 16 GB runs up to about 20B (including gpt-oss 20B), 24 GB runs 30 to 32B, and 70B models need about 48 GB or a large-memory Mac.

Should I buy a GPU or a Mac?

A GPU is faster per dollar for models up to about 32B. A Mac with 64 to 128 GB of unified memory is the simplest way to run 70B-class and large mixture-of-experts models, at lower speed.