GPUs with 40 GB or more VRAM for Running LLMs
Browse 27 GPUs with 40 GB or more VRAM compatible with running LLM models locally. Compare VRAM, memory bandwidth, and AI performance. Top picks and budget guide
27 GPUsClear filters
| GPU | Price |
|---|---|
| RTX PRO 6000 BlackwellPro96 GBGDDR7 · 1,792 GB/s · 600 WNVIDIA · Blackwell · 2025 | about $8,565 |
| H200 SXMPro141 GBHBM3e · 4,800 GB/s · 700 WNVIDIA · Hopper · 2024 | – |
| B200Pro192 GBHBM3e · 8,000 GB/s · 1000 WNVIDIA · Blackwell · 2024 | – |
| H100 SXMPro80 GBHBM3 · 3,352 GB/s · 700 WNVIDIA · Hopper · 2022 | – |
| H100 PCIePro80 GBHBM2e · 2,039 GB/s · 350 WNVIDIA · Hopper · 2022 | – |
| Arc Pro B60 Dual 48GBPro48 GBGDDR6 · 456 GB/s · 400 WIntel · Xe2 · 2025 | about $1,200 |
| A100 80GB SXMPro80 GBHBM2e · 2,039 GB/s · 400 WNVIDIA · Ampere · 2020 | – |
| RTX PRO 5000 BlackwellPro72 GBGDDR7 · 1,344 GB/s · 300 WNVIDIA · Blackwell · 2025 | about $4,500 |
| RTX PRO 6000 Blackwell Max-QPro96 GBGDDR7 · 1,792 GB/s · 300 WNVIDIA · Blackwell · 2025 | about $8,565 |
| A100 40GB PCIePro40 GBHBM2e · 1,555 GB/s · 250 WNVIDIA · Ampere · 2020 | – |
| H200 NVLPro141 GBHBM3e · 4,800 GB/s · 600 WNVIDIA · Hopper · 2024 | – |
| B300Pro288 GBHBM3e · 8,000 GB/s · 1400 WNVIDIA · Blackwell Ultra · 2025 | – |
| Instinct MI300XPro192 GBHBM3 · 5,300 GB/s · 750 WAMD · CDNA 3 · 2023 | – |
| Instinct MI325XPro256 GBHBM3e · 6,000 GB/s · 1000 WAMD · CDNA 3 · 2024 | – |
| Instinct MI350XPro288 GBHBM3e · 8,000 GB/s · 1000 WAMD · CDNA 4 · 2025 | – |
| Instinct MI355XPro288 GBHBM3e · 8,000 GB/s · 1400 WAMD · CDNA 4 · 2025 | – |
| GH200 Grace Hopper SuperchipPro144 GBHBM3e · 4,900 GB/s · 1000 WNVIDIA · Hopper (Grace Hopper) · 2024 | – |
| L40SPro48 GBGDDR6 · 864 GB/s · 350 WNVIDIA · Ada Lovelace · 2023 | – |
| RTX PRO 6000 Blackwell ServerPro96 GBGDDR7 · 1,597 GB/s · 600 WNVIDIA · Blackwell · 2025 | – |
| Quadro RTX 8000Pro48 GBGDDR6 · 672 GB/s · 260 WNVIDIA · Turing · 2018 | about $9,999 |
| L40Pro48 GBGDDR6 · 864 GB/s · 300 WNVIDIA · Ada Lovelace · 2022 | – |
| RTX 6000 AdaPro48 GBGDDR6 · 960 GB/s · 300 WNVIDIA · Ada Lovelace · 2022 | about $6,799 |
| Instinct MI250XPro128 GBHBM2e · 3,277 GB/s · 560 WAMD · CDNA 2 · 2021 | – |
| Radeon PRO W7900Pro48 GBGDDR6 · 864 GB/s · 295 WAMD · RDNA 3 · 2023 | about $3,999 |
| A40Pro48 GBGDDR6 · 696 GB/s · 300 WNVIDIA · Ampere · 2020 | – |
Which GPU Do You Need for AI?
The amount of VRAM is the most important specification for running LLMs locally. Most 7B parameter models require 4–8 GB of VRAM at common quantization levels, while 70B models need 24–48 GB. Memory bandwidth determines how fast the model generates tokens — faster bandwidth means faster responses.