What is the best GPU under $1,000 for local LLMs?

By Q4/8K decode speed on Llama-3.1-8B-Instruct: the RTX 5080 16GB at about 136.2 tok/s (measured), with 16 GB VRAM and 960 GB/s bandwidth at a $999 MSRP.

How large a model can a sub-$1,000 GPU run?

Across the 30 GPUs with an MSRP under $1,000, the largest catalog model any of them fully fits at Q4/8K is Mixtral-8x7B-Instruct-v0.1 (46.7B) on the Mac mini M4 32GB. Anything larger needs more VRAM or partial offload at much lower speed.

Best GPU under $1,000 for local LLMs (Q4/8K)

30 catalog GPUs carry an MSRP under $1,000. They are ranked by decode speed (Q4 / 8K context) on Llama-3.1-8B-Instruct — a representative, widely used 8B-class model. MSRP is the launch list price, not a live price: street and Amazon prices move, so click through for the current number. Methodology.

#	GPU	MSRP	VRAM (GB)	Bandwidth (GB/s)	Llama-3.1-8B-Instruct tok/s	Fits?	Buy
1	RTX 5080 16GB	$999	16	960	136.2 measured	Yes	NVIDIA GeForce RTX 5080 Founders Edition (affiliate) · More listings
2	RTX 5070 Ti 16GB	$749	16	896	131.9 measured	Yes	Search Amazon (affiliate)
3	RTX 5070 12GB	$549	12	672	100.4 measured	Yes	Search Amazon (affiliate)
4	RTX 4070 Ti Super 16GB	$799	16	672	95.0 measured	Yes	Search Amazon (affiliate)
5	RX 9070 XT 16GB	$599	16	640	89.1 measured	Yes	Search Amazon (affiliate)
6	RX 9070 16GB	$549	16	640	87.6 measured	Yes	Search Amazon (affiliate)
7	RTX 5060 Ti 16GB	$429	16	448	74.7 measured	Yes	Search Amazon (affiliate)
8	RTX 3080 10GB	$699	10	760	70.0 measured	Yes	Search Amazon (affiliate)
9	RX 7800 XT 16GB	$499	16	624	68.0 measured	Yes	Search Amazon (affiliate)
10	RX 7900 XTX 24GB	$999	24	960	65.0 measured	Yes	ASRock Radeon RX 7900 XTX Phantom Gaming 24GB OC (affiliate) · More listings
11	RX 7900 XT 20GB	$899	20	800	63.6 est.	Yes	Search Amazon (affiliate)
12	RTX 4080 Super 16GB	$999	16	736	58.0 measured	Yes	ASUS ProArt GeForce RTX 4080 Super OC (affiliate) · More listings
13	RTX 4070 12GB	$599	12	504	52.1 measured	Yes	ASUS Dual GeForce RTX 4070 OC 12GB (affiliate) · More listings
14	RTX 4060 Ti 16GB	$499	16	288	50.0 measured	Yes	Search Amazon (affiliate)
15	RX 7900 GRE 16GB	$549	16	576	45.8 est.	Yes	Search Amazon (affiliate)
16	Arc A770 16GB	$349	16	560	44.5 est.	Yes	Search Amazon (affiliate)
17	RTX 4070 Ti 12GB	$799	12	554	44.1 est.	Yes	Search Amazon (affiliate)
18	RTX 3060 12GB	$329	12	360	42.0 measured	Yes	MSI GeForce RTX 3060 Ventus 2X 12G OC (affiliate) · More listings
19	RTX 4070 Super 12GB	$599	12	504	40.1 est.	Yes	Search Amazon (affiliate)
20	Arc B580 12GB	$249	12	456	36.3 est.	Yes	Search Amazon (affiliate)
21	RTX 3070 8GB	$499	8	448	35.6 est.	Yes	Search Amazon (affiliate)
22	RTX 5060 8GB	$299	8	448	35.6 est.	Yes	Search Amazon (affiliate)
23	RTX 5060 Ti 8GB	$379	8	448	35.6 est.	Yes	Search Amazon (affiliate)
24	RX 7700 XT 12GB	$449	12	432	34.4 est.	Yes	Search Amazon (affiliate)
25	Arc B570 10GB	$219	10	380	30.2 est.	Yes	Search Amazon (affiliate)
26	RTX 4060 Ti 8GB	$399	8	288	22.9 est.	Yes	Search Amazon (affiliate)
27	RX 7600 8GB	$269	8	288	22.9 est.	Yes	Search Amazon (affiliate)
28	RX 7600 XT 16GB	$329	16	288	22.9 est.	Yes	Search Amazon (affiliate)
29	RTX 4060 8GB	$299	8	272	21.6 est.	Yes	Search Amazon (affiliate)
30	Mac mini M4 32GB	$999	32	120	21.0 measured	Yes	Search Amazon (affiliate)

tok/s figures are bandwidth-based estimates unless a row is tagged measured. The Fits? column links to the full Llama-3.1-8B-Instruct fit check on each card.

More lists: 7B-class models · 13B-class models · 30B-class models · 70B+ models · Under $500 · Under $2,000 · 8 GB VRAM · 12 GB VRAM · 16 GB VRAM · 24 GB VRAM · For Qwen3 · For DeepSeek · For Llama · For Gemma · For GLM · For coding · For reasoning · MoE picks

Best GPUs under $1,000 for local LLMs