Guide · Local AI

Choose a GPU for local AI

Latest price sample: 2026-11-20 · Hardware advice reviewed: 2026-10-04 · Methodology

Start with the workload, then compare the price. The RTX 3090 and RTX 4090 both have 24GB. The RTX 5090 has 32GB, while workstation cards can offer 48GB or 96GB. More memory is useful only if your workload needs it; software support, power, and total build cost also matter.

Use the dated asking-price samples below when comparing listings. Our older editorial fair values are not current offers or guaranteed buying targets.

A shorter path to your next GPU

Narrow your GPU shortlist

Set your card budget and required memory. Compare up to three options with the lowest qualifying sampled prices.

Use a memory requirement that includes context and working buffers. NVIDIA alone does not guarantee support for your application, model or quantization.

Choose software and memory together

Check the model architecture, quantization, operating system and driver supported by your runtime. Ollama lists different AMD support on Linux and Windows; NVIDIA and Apple have their own requirements. See Ollama hardware documentation.

Estimate weights plus context cache and runtime working memory using our VRAM guide. More capacity helps only when the software can use it. We have not benchmarked the cards below; this is a buying shortlist, not a speed ranking.

Browse all 45 GPU options and sampled prices

Models and sampled asking prices

These are filtered eBay Buy It Now asking prices, not completed sales, offers, or a performance ranking. Shipping, tax, seller terms, and condition can change the comparison.

Paid links: we may earn a commission from eBay or Amazon purchases. Full disclosure.

NVIDIA GeForce RTX 5090

32GB GDDR7 · 575W reference TDP

$6,899 median asking price · 192 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 3090

24GB GDDR6X · 350W reference TDP

$2,000 median asking price · 116 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4090

24GB GDDR6X · 450W reference TDP

$3,600 median asking price · 77 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 5080

16GB GDDR7 · 360W reference TDP

$2,200 median asking price · 158 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 5070 Ti

16GB GDDR7 · 300W reference TDP

$1,500 median asking price · 85 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 5070

12GB GDDR7 · 250W reference TDP

$950 median asking price · 76 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 5060 Ti 16GB

16GB GDDR7 · 180W reference TDP

$890 median asking price · 50 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 5060 Ti 8GB

8GB GDDR7 · 180W reference TDP

$590 median asking price · 42 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 5060

8GB GDDR7 · 145W reference TDP

$534 median asking price · 101 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4070 Ti Super

16GB GDDR6X · 285W reference TDP

$1,300 median asking price · 34 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4080

16GB GDDR6X · 320W reference TDP

$1,450 median asking price · 48 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4080 Super

16GB GDDR6X · 320W reference TDP

$1,750 median asking price · 28 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4070

12GB GDDR6X · 200W reference TDP

$790 median asking price · 52 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 3080 10GB

10GB GDDR6X · 320W reference TDP

$520 median asking price · 142 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 3080 12GB

12GB GDDR6X · 350W reference TDP

No fresh, sufficiently sized sample available for this model.

Specs and buying checks →

NVIDIA GeForce RTX 3060 12GB

12GB GDDR6 · 170W reference TDP

$400 median asking price · 88 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX A6000

48GB GDDR6 · 300W reference TDP

$5,309 median asking price · 24 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX 6000 Ada

48GB GDDR6 · 300W reference TDP

No fresh, sufficiently sized sample available for this model.

Specs and buying checks →

AMD Radeon RX 7900 XTX

24GB GDDR6 · 355W reference TDP

$1,440 median asking price · 18 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon RX 7900 XT

20GB GDDR6 · 300W reference TDP

$970 median asking price · 11 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon RX 7800 XT

16GB GDDR6 · 263W reference TDP

$725 median asking price · 15 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon RX 9070 XT

16GB GDDR6 · 304W reference TDP

$1,100 median asking price · 85 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon RX 9070

16GB GDDR6 · 220W reference TDP

$850 median asking price · 11 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4060

8GB GDDR6 · 115W reference TDP

$420 median asking price · 88 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 3070

8GB GDDR6 · 220W reference TDP

$380 median asking price · 149 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4070 Super

12GB GDDR6X · 220W reference TDP

$950 median asking price · 44 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA GeForce RTX 4060 Ti 16GB

16GB GDDR6 · 165W reference TDP

$740 median asking price · 11 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon RX 6700 XT

12GB GDDR6 · 230W reference TDP

$450 median asking price · 62 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon RX 6800 XT

16GB GDDR6 · 300W reference TDP

$625 median asking price · 24 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA Tesla P40

24GB GDDR5 · 250W reference TDP

$396 median asking price · 6 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon RX 9060 XT 16GB

16GB GDDR6 · 160W reference TDP

$785 median asking price · 27 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX PRO 6000 Blackwell Workstation Edition

96GB GDDR7 ECC · 600W reference TDP

$19,000 median asking price · 9 valid samples · 2026-11-20

Specs and buying checks →

Intel Arc B580

12GB GDDR6 · 190W reference TDP

No fresh, sufficiently sized sample available for this model.

Specs and buying checks →

Intel Arc B570

10GB GDDR6 · 150W reference TDP

No fresh, sufficiently sized sample available for this model.

Specs and buying checks →

Intel Arc Pro B50

16GB GDDR6 · 70W reference TDP

$628 median asking price · 9 valid samples · 2026-11-20

Specs and buying checks →

Intel Arc Pro B60

24GB GDDR6 · 200W reference TDP

$1,000 median asking price · 8 valid samples · 2026-11-20

Specs and buying checks →

Intel Arc Pro B65

32GB GDDR6 · 200W reference TDP

No fresh, sufficiently sized sample available for this model.

Specs and buying checks →

Intel Arc Pro B70

32GB GDDR6 · 290W reference TDP

$2,174 median asking price · 9 valid samples · 2026-11-20

Specs and buying checks →

AMD Radeon AI PRO R9700

32GB GDDR6 · 300W reference TDP

$2,300 median asking price · 9 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX PRO 5000 Blackwell 48GB

48GB GDDR7 ECC · 300W reference TDP

$7,290 median asking price · 8 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX PRO 5000 Blackwell 72GB

72GB GDDR7 ECC · 300W reference TDP

$9,500 median asking price · 7 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX PRO 4500 Blackwell Workstation Edition

32GB GDDR7 ECC · 200W reference TDP

$4,499 median asking price · 13 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX PRO 4000 Blackwell

24GB GDDR7 ECC · 145W reference TDP

$2,980 median asking price · 19 valid samples · 2026-11-20

Specs and buying checks →

NVIDIA RTX PRO 4000 Blackwell SFF Edition

24GB GDDR7 ECC · 70W reference TDP

No fresh, sufficiently sized sample available for this model.

Specs and buying checks →

NVIDIA RTX PRO 2000 Blackwell

16GB GDDR7 ECC · 70W reference TDP

$1,500 median asking price · 10 valid samples · 2026-11-20

Specs and buying checks →

Used versus new

Compare the total checkout cost with alternatives that meet your memory and software requirements. An older editorial estimate is context, not a price you can assume is available. Check the condition, return policy, and any warranty offered for each listing.

Macs and the rest of the PC

Mac unified memory is shared with the operating system and applications. Compare available memory and matched workload results, rather than assuming a Mac memory figure equals dedicated VRAM. See Mac vs GPU.

For a PC, size system RAM and storage for the actual artifacts and loading or offloading behavior. Include the host system, PSU, cooling and case clearance in the budget. Follow the board manufacturer's power and connector requirements.

Buying checklist

Sources and review

Hardware advice reviewed October 4, 2026. These are documentation-based recommendations; we have not benchmarked these configurations.

FAQ

What is the best used GPU for local AI in 2026?

There is no single best used GPU for every local AI workload. Start with the memory your model and context need, check software support, then compare dated asking prices and the cost of a compatible build. The RTX 3090 and RTX 4090 both offer 24GB; the RTX 5090 offers 32GB; workstation options include 48GB cards and the 96GB RTX PRO 6000 Blackwell.

What local LLM hardware do I need to get started?

Start with the exact model artifact, runtime, context and concurrency. Check available GPU and system memory, storage for the artifacts, and power and cooling requirements. Small models can run on hardware you already own; there is no universal minimum build.

How much VRAM do I need for local LLMs?

Weights are only part of the requirement. Add context cache and runtime working memory, and confirm support for the exact quantization. Use our VRAM guide to estimate capacity rather than treating parameter counts as guaranteed fits.

Always-on agents that call hosted models

A different job than local inference: bots and automations that stay up and call an API. For that, a cheap Linux VPS is the usual box. VPSDime is the referral we use when that is the workload. (Affiliate link — see disclosure.)

See RTX 3090 price guideVRAM guideMuse Glimmer hardwareAffiliate disclosure