RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?API pricesVRAM calculatorNew

Best GPU for running LLMs locally under $800

For local models, VRAM decides what runs and memory bandwidth decides how fast. These are the 23 consumer GPUs in our table with a US launch MSRP of $800 or less, scored on both with the same formulas as the rest of the site. Prices are launch MSRPs, not today's street prices, which can be higher or lower.

RTX 5070 Timost VRAM (16 GB, $749)
RTX 5070 Tifastest (896 GB/s, $749)
167models at 4-bit+ on the Arc A770 16GB
Arc B580cheapest listed ($249)

The short answer

Every card under $800

Sorted by VRAM, then bandwidth. Model columns show the best quant that fits fully in VRAM at 8K context and an estimated speed ceiling, or the llama.cpp offload plan when the model only runs with part of it in system RAM (speed range from DDR4-3200 to DDR5-6000). How we calculate · offload method.

GPULaunch MSRPVRAMBandwidthModels at 4-bit+Qwen3 8Bgpt-oss 20BQwen3 32Bgpt-oss 120B
RTX 5070 Ti$74916 GB896 GB/s167Q8_0 · ~103 tok/sQ8_0 · ~369 tok/sOffload Q4_K_M, 32 GB RAM · ~3.3–6.4 tok/sOffload Q4_K_M, 64 GB RAM · ~14–29 tok/s
RTX 4070 Ti Super$79916 GB672 GB/s167Q8_0 · ~77 tok/sQ8_0 · ~277 tok/sOffload Q4_K_M, 32 GB RAM · ~3.2–6.1 tok/sOffload Q4_K_M, 64 GB RAM · ~14–28 tok/s
RX 9070$54916 GB640 GB/s167Q8_0 · ~73 tok/sQ8_0 · ~264 tok/sOffload Q4_K_M, 32 GB RAM · ~3.2–6.1 tok/sOffload Q4_K_M, 64 GB RAM · ~14–28 tok/s
RX 9070 XT$59916 GB640 GB/s167Q8_0 · ~73 tok/sQ8_0 · ~264 tok/sOffload Q4_K_M, 32 GB RAM · ~3.2–6.1 tok/sOffload Q4_K_M, 64 GB RAM · ~14–28 tok/s
RX 7800 XT$49916 GB624 GB/s167Q8_0 · ~72 tok/sQ8_0 · ~257 tok/sOffload Q4_K_M, 32 GB RAM · ~3.2–6.0 tok/sOffload Q4_K_M, 64 GB RAM · ~14–28 tok/s
RX 7900 GRE$54916 GB576 GB/s167Q8_0 · ~66 tok/sQ8_0 · ~237 tok/sOffload Q4_K_M, 32 GB RAM · ~3.2–6.0 tok/sOffload Q4_K_M, 64 GB RAM · ~14–28 tok/s
Arc A770 16GB$34916 GB560 GB/s167Q8_0 · ~64 tok/sQ8_0 · ~231 tok/sOffload Q4_K_M, 32 GB RAM · ~3.2–5.9 tok/sOffload Q4_K_M, 64 GB RAM · ~14–27 tok/s
RTX 5060 Ti 16GB$42916 GB448 GB/s167Q8_0 · ~51 tok/sQ8_0 · ~185 tok/sOffload Q4_K_M, 32 GB RAM · ~3.1–5.6 tok/sOffload Q4_K_M, 64 GB RAM · ~14–27 tok/s
RTX 4060 Ti 16GB$49916 GB288 GB/s167Q8_0 · ~33 tok/sQ8_0 · ~119 tok/sOffload Q4_K_M, 32 GB RAM · ~2.9–5.0 tok/sOffload Q4_K_M, 64 GB RAM · ~13–25 tok/s
RTX 5070$54912 GB672 GB/s150Q8_0 · ~77 tok/sOffload Q4_K_M, 32 GB RAM · ~85–124 tok/sOffload Q4_K_M, 32 GB RAM · ~2.2–4.4 tok/sOffload Q4_K_M, 64 GB RAM · ~13–27 tok/s
RTX 4070$59912 GB504 GB/s150Q8_0 · ~58 tok/sOffload Q4_K_M, 32 GB RAM · ~75–104 tok/sOffload Q4_K_M, 32 GB RAM · ~2.2–4.3 tok/sOffload Q4_K_M, 64 GB RAM · ~13–26 tok/s
RTX 4070 Super$59912 GB504 GB/s150Q8_0 · ~58 tok/sOffload Q4_K_M, 32 GB RAM · ~75–104 tok/sOffload Q4_K_M, 32 GB RAM · ~2.2–4.3 tok/sOffload Q4_K_M, 64 GB RAM · ~13–26 tok/s
RTX 4070 Ti$79912 GB504 GB/s150Q8_0 · ~58 tok/sOffload Q4_K_M, 32 GB RAM · ~75–104 tok/sOffload Q4_K_M, 32 GB RAM · ~2.2–4.3 tok/sOffload Q4_K_M, 64 GB RAM · ~13–26 tok/s
Arc B580$24912 GB456 GB/s150Q8_0 · ~52 tok/sOffload Q4_K_M, 32 GB RAM · ~72–98 tok/sOffload Q4_K_M, 32 GB RAM · ~2.2–4.2 tok/sOffload Q4_K_M, 64 GB RAM · ~13–25 tok/s
RX 7700 XT$44912 GB432 GB/s150Q8_0 · ~50 tok/sOffload Q4_K_M, 32 GB RAM · ~70–94 tok/sOffload Q4_K_M, 32 GB RAM · ~2.2–4.2 tok/sOffload Q4_K_M, 64 GB RAM · ~13–25 tok/s
RTX 3060 12GB$32912 GB360 GB/s150Q8_0 · ~41 tok/sOffload Q4_K_M, 32 GB RAM · ~63–83 tok/sOffload Q4_K_M, 32 GB RAM · ~2.1–4.1 tok/sOffload Q4_K_M, 64 GB RAM · ~13–25 tok/s
RTX 3080 10GB$69910 GB760 GB/s146Q6_K · ~113 tok/sOffload Q4_K_M, 32 GB RAM · ~49–85 tok/sOffload Q4_K_M, 32 GB RAM · ~1.9–3.9 tok/sOffload Q4_K_M, 64 GB RAM · ~13–26 tok/s
RTX 5060$2998 GB448 GB/s135Q4_K_M · ~89 tok/sOffload Q4_K_M, 32 GB RAM · ~29–52 tok/sOffload Q4_K_M, 32 GB RAM · ~1.6–3.3 tok/sOffload Q4_K_M, 64 GB RAM · ~12–24 tok/s
RTX 3060 Ti$3998 GB448 GB/s135Q4_K_M · ~89 tok/sOffload Q4_K_M, 32 GB RAM · ~29–52 tok/sOffload Q4_K_M, 32 GB RAM · ~1.6–3.3 tok/sOffload Q4_K_M, 64 GB RAM · ~12–24 tok/s
RTX 3070$4998 GB448 GB/s135Q4_K_M · ~89 tok/sOffload Q4_K_M, 32 GB RAM · ~29–52 tok/sOffload Q4_K_M, 32 GB RAM · ~1.6–3.3 tok/sOffload Q4_K_M, 64 GB RAM · ~12–24 tok/s
RX 7600$2698 GB288 GB/s135Q4_K_M · ~57 tok/sOffload Q4_K_M, 32 GB RAM · ~27–45 tok/sOffload Q4_K_M, 32 GB RAM · ~1.6–3.2 tok/sOffload Q4_K_M, 64 GB RAM · ~12–23 tok/s
RTX 4060 Ti 8GB$3998 GB288 GB/s135Q4_K_M · ~57 tok/sOffload Q4_K_M, 32 GB RAM · ~27–45 tok/sOffload Q4_K_M, 32 GB RAM · ~1.6–3.2 tok/sOffload Q4_K_M, 64 GB RAM · ~12–23 tok/s
RTX 4060$2998 GB272 GB/s135Q4_K_M · ~54 tok/sOffload Q4_K_M, 32 GB RAM · ~27–44 tok/sOffload Q4_K_M, 32 GB RAM · ~1.6–3.2 tok/sOffload Q4_K_M, 64 GB RAM · ~12–23 tok/s

How to choose

Start from the model you want. A model needs roughly its 4-bit file size plus KV cache and about 1 GB of runtime overhead. Pick the cheapest card where it fits with room for your context length, then prefer higher bandwidth for faster replies. If a model is just out of reach, mixture-of-experts models such as gpt-oss and Qwen3 A3B can still run with most expert weights in system RAM at a usable speed; dense 70B models usually cannot.

By memory size: Best models for 8 GBBest models for 12 GBBest models for 16 GBBest models for 24 GBBest models for 32 GB

Other budgets: Under $300Under $500Under $1,000Under $1,500Under $2,000 · Mac vs PC · Can I run…?

Prices: NVIDIA, AMD and Intel US launch MSRPs (Founders/reference editions). Specs: vendor spec pages, linked on each GPU page. Speed is a bandwidth ceiling, not a benchmark; real speed is typically 50–80% of it. Data refreshes hourly. Report an error.