AI models that fit on a 80 GB GPU

Open models that fit in 80 GB of VRAM: 164 as published, 28 at FP8 and 17 at INT4. Each model is listed on the smallest tier it fits at that precision, so anything smaller is on the 48 GB page or below.

GPUs with 80 GB

Cards whose datasheet VRAM puts them in this tier, up to the next tier at 96 GB. Prices are the lowest live data-center on-demand rate per GPU.

GPUVRAMFrom per GPU hour
A10080 GB$1.49/hr
H10080 GB$3.00/hr
H100 NVL94 GBNo live offer

Fits as published

Models whose published weights, plus the flat overhead, fit this much memory with no quantization.

Text models

The 40 most downloaded of 93.

ModelParametersPublished asVRAM needed
OTel-2.0-LLM-31B-IT32.1BBF1671.8 GB
Qwen3-32B32.8BBF1673.2 GB
dolphin-2.9.1-yi-1.5-34b34.4BBF1676.9 GB
Qwen3-30B-A3B30.5BBF1668.2 GB
Qwen2.5-32B-Instruct32.8BBF1673.2 GB
GLM-4.7-Flash31.2BBF1669.8 GB
Qwen2.5-Coder-32B-Instruct32.8BBF1673.2 GB
Qwen3-30B-A3B-Instruct-250730.5BBF1668.2 GB
Qwen3.6-35B-A3B-abliterated-v434.7BBF1677.5 GB
NVIDIA-Nemotron-3-Nano-30B-A3B-BF1631.6BBF1670.6 GB
Qwen3.8-27B-OBLITERATED27.8BBF1662.1 GB
Qwen3-Coder-30B-A3B-Instruct30.5BBF1668.2 GB
DeepSeek-R1-Distill-Qwen-32B32.8BBF1673.2 GB
Dolphin-Mistral-24B-Venice-Edition24.0BBF1653.7 GB
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF1631.6BBF1670.6 GB
Ornith-1.0-35B-AEON-Ultimate-Uncensored-BF1635.1BBF1678.5 GB
granite-4.1-30b28.9BBF1664.5 GB
automotive34.7BBF1677.5 GB
Llama-3_3-Nemotron-Super-49B-v1_5-FP849.9BF8_E4M355.7 GB
droplychee-1.0-27b27.8BBF1662.1 GB
gemma-4-31B-it-uncensored32.7BBF1673.1 GB
Ornith-1.5-35B-A3B-MLX34.7BBF1677.5 GB
Meta-Llama-3.1-70B-Instruct-FP870.6BF8_E4M378.8 GB
Qwen3-30B-A3B-Base30.5BBF1668.2 GB
NVIDIA-Nemotron-3-Nano-30B-A3B-Base-BF1631.6BBF1670.6 GB
Llama-3.3-70B-Instruct-FP8-dynamic70.6BF8_E4M378.9 GB
lynx-instruct-30b30.5BBF1668.2 GB
Goedel-Prover-V2-32B32.8BBF1673.2 GB
Llama-3.1-70B-Instruct-FP870.6BF8_E4M378.8 GB
Qwen2.5-32B32.8BBF1673.2 GB
Qwen3-30B-A3B-Thinking-250730.5BBF1668.2 GB
Qwen-AgentWorld-35B-A3B34.7BBF1677.5 GB
QwQ-32B32.8BBF1673.2 GB
Laguna-XS-2.133.4BBF1674.8 GB
Xing4.0-29B-A4B31.2BBF1669.8 GB
Nemotron-Cascade-2-30B-A3B31.6BBF1670.6 GB
Tongyi-DeepResearch-30B-A3B30.5BBF1668.2 GB
KAT-Coder-V2.5-Dev34.7BBF1677.5 GB
Olmo-3-1125-32B32.2BBF1672.0 GB
gemma-2-27b-it27.2BBF1660.9 GB

Vision-language models

31 models.

ModelParametersPublished asVRAM needed
gemma-4-31B-it31.3BBF1669.9 GB
gemma-4-26B-A4B-it25.8BBF1657.7 GB
Qwen3.6-27B27.8BBF1662.1 GB
Qwen3.8-27B27.8BBF1662.1 GB
Qwen3.5-27B27.8BBF1662.1 GB
JEV-27B-VL27.8BBF1662.1 GB
Qwen2.5-VL-32B-Instruct33.5BBF1674.8 GB
diffusiongemma-26B-A4B-it25.8BBF1657.7 GB
gemma-4-31B32.7BBF1673.1 GB
Muse-Glimmer-30B29.8BBF1666.6 GB
Qwen3-VL-32B-Instruct33.4BBF1674.6 GB
gemma-3-27b-it27.4BBF1661.3 GB
Infinity-Parser2-Pro35.1BBF1678.5 GB
Qwen3-VL-30B-A3B-Instruct31.1BBF1669.4 GB
medgemma-27b-it28.8BBF1664.5 GB
gemma-4-26B-A4B-it-qat-q4_0-unquantized26.5BBF1659.3 GB
InternVL2-26B25.5BBF1657.0 GB
Huihui-Qwen3.5-27B-abliterated27.8BBF1662.1 GB
Qwen3.8-27B-Uncensored27.8BBF1662.1 GB
Huihui-Qwen3.8-27B-abliterated27.8BBF1662.1 GB
Qwen3.8-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS27.4BBF1661.1 GB
Qwen-Image-Bench27.4BBF1661.1 GB
clef27.4BBF1661.1 GB
Swift-1.5-Qwen3.8-27b27.8BBF1662.1 GB
ThinkingCap-Qwen3.8-27B27.8BBF1662.1 GB
AREX-227.4BBF1661.1 GB
SenseNova-6.8-Code2Video27.4BBF1661.1 GB
Thomson-1.0-Small35.1BBF1678.5 GB
Holo4-27B27.4BBF1661.1 GB
Thinklite-27B27.8BBF1662.1 GB
Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU27.8BBF1662.1 GB

Image generation models

2 models.

ModelParametersPublished asVRAM needed
minimax-h3-image-pdmd-2step33.1BBF1674.0 GB
P-H3-Image33.1BBF1674.0 GB

Video generation models

14 models.

ModelParametersPublished asVRAM needed
Wan2.2-T2V-A14B-Diffusers14.3BF3263.9 GB
Wan2.2-I2V-A14B-Diffusers14.3BF3263.9 GB
Wan2.1-I2V-14B-720P-Diffusers16.4BF3273.3 GB
Wan2.1-T2V-14B-Diffusers14.3BF3263.9 GB
Wan2.1-I2V-14B-480P-Diffusers16.4BF3273.3 GB
Wan2.1-I2V-14B-480P16.4BF3273.3 GB
Wan2.1-T2V-14B14.3BF3263.9 GB
Wan2.1-I2V-14B-720P16.4BF3273.3 GB
Wan2.1-VACE-14B17.3BF3277.5 GB
Pika-H3-Flex33.1BBF1674.0 GB
FastVideo-Minimax-FastH3-Preview-v0.235.0BBF1678.3 GB
Kandinsky-6.0-Pro-distill-5s-Diffusers30.1BBF1667.4 GB
Kandinsky-6.0-Pro-5s-Diffusers30.1BBF1667.4 GB
FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree35.0BBF1678.3 GB

Speech models

2 models.

ModelParametersPublished asVRAM needed
NVIDIA-Nemotron-3.5-Transcribe-30B-A4B33.7BBF1675.3 GB
SOV-Transcribe-126.6BBF1659.5 GB

Embedding models

1 model.

ModelParametersPublished asVRAM needed
meta-encoder-v128.4BBF1663.6 GB

Other models

21 models.

ModelParametersPublished asVRAM needed
MiniMax-H333.1BBF1674.0 GB
GEV-26B-Decide25.8BBF1657.7 GB
FLUX.2-dev32.2BBF1672.0 GB
JEV-27B26.9BBF1660.1 GB
AutoTrust-Decider-Pro-Vision-27B-v1.127.4BBF1661.1 GB
AutoTrust-JEV-27B-v1.127.4BBF1661.1 GB
MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-Ref2VA27.4BBF1661.1 GB
Mistral-Small-24B-Instruct-250123.6BBF1652.7 GB
Qwen3.8-Whittle-MoE-27B-A17.8B26.9BBF1660.2 GB
openjev27.4BBF1661.1 GB
mammoth-127.4BBF1661.1 GB
rune-26b-a4b-GGUF25.8BBF1657.7 GB
NVIDIA-NemotronLabs-VoiceChat-11B11.1BF3249.6 GB
pplx-decider-v1-27b26.1BBF1658.3 GB
Artemis-31B-v1.231.3BBF1669.9 GB
kev-27b25.6BBF1657.3 GB
Apriel-Voice-27B-preview27.2BBF1660.9 GB
Orion-26B-A4B-v1.125.8BBF1657.7 GB
pplx-decider-v1.326.1BBF1658.3 GB
PAN229.2BBF1665.3 GB
Orion-26B-A4B-v125.8BBF1657.7 GB

Fits at FP8

Models that fit only after quantizing the weights to 8 bits (1 byte per parameter). Needs an FP8 checkpoint or an engine that quantizes on load, and a GPU with FP8 support.

Text models

22 models.

ModelParametersPublished asVRAM needed
Llama-3.3-70B-Instruct70.6BBF1678.8 GB
Llama-3.1-70B-Instruct70.6BBF1678.8 GB
Kimi-Linear-48B-A3B-Instruct49.1BBF1654.9 GB
Llama-3_3-Nemotron-Super-49B-v149.9BBF1655.7 GB
Meta-Llama-3-70B70.6BBF1678.8 GB
DeepSeek-R1-Distill-Llama-70B70.6BBF1678.8 GB
Llama-3.1-70B-LatamGPT-SFT-1.070.6BBF1678.8 GB
Meta-Llama-3-70B-Instruct70.6BBF1678.8 GB
Llama-3.1-70B70.6BBF1678.8 GB
Meta-Llama-3.1-70B70.6BBF1678.8 GB
Meta-Llama-3.1-70B-Instruct70.6BBF1678.8 GB
Apertus-70B-Instruct-250970.6BBF1678.9 GB
Nous-Hermes-2-Mixtral-8x7B-DPO46.7BBF1652.2 GB
LongCat-Flash-Lite69.1BBF1677.2 GB
Llama-3_3-Nemotron-Super-49B-v1_549.9BBF1655.7 GB
Nemotron-H-56B-Base-8K56.3BBF1662.9 GB
Qwen2-57B-A14B-Instruct57.4BBF1664.2 GB
Meta-Llama-3.1-70B-Instruct70.6BBF1678.8 GB
Llama-3.1-Nemotron-70B-Instruct-HF70.6BBF1678.8 GB
Jamba-v0.151.6BBF1657.6 GB
Qwen2-57B-A14B57.4BBF1664.2 GB
Hermes-3-Llama-3.1-70B70.6BBF1678.8 GB

Vision-language models

1 model.

ModelParametersPublished asVRAM needed
Qwen3.8-Flash-Next-Uncensored-MLX71.3BBF1679.7 GB

Image generation models

1 model.

ModelParametersPublished asVRAM needed
Cosmos3-Super-Text2Image-4Step64.0BBF1671.5 GB

Video generation models

1 model.

ModelParametersPublished asVRAM needed
Cosmos3-Super-Image2Video64.6BBF1672.2 GB

Other models

3 models.

ModelParametersPublished asVRAM needed
Mixtral-8x7B-Instruct-v0.146.7BBF1652.2 GB
Ripple-65B-Preview64.8BBF1672.4 GB
Ripple-65B-Base-Preview64.8BBF1672.4 GB

Fits at INT4

Models that fit only after quantizing to 4 bits (0.5 byte per parameter). Needs a quantized checkpoint actually published for the model; check its Hugging Face page before relying on a row.

Text models

11 models.

ModelParametersPublished asVRAM needed
NVIDIA-Nemotron-3-Super-120B-A12B-BF16123.6BBF1669.1 GB
NVIDIA-Nemotron-3-Super-120B-A12B-FP8123.6BF8_E4M369.1 GB
GLM-4.5-Air110.5BBF1661.7 GB
GLM-4.5-Air-FP8110.5BF8_E4M361.8 GB
Laguna-S-2.1-FP8117.6BF8_E4M365.7 GB
Laguna-S-2.1117.6BBF1665.7 GB
Ling-3.0-flash127.5BBF1671.2 GB
xLAM-8x22b-r140.6BBF1678.6 GB
Solar-Open-100B102.7BBF1657.4 GB
NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16123.6BBF1669.1 GB
GLM-4.5-Air-Base110.5BBF1661.7 GB

Vision-language models

5 models.

ModelParametersPublished asVRAM needed
Qwen3.5-122B-A10B-FP8125.1BF8_E4M369.9 GB
Qwen3.5-122B-A10B125.1BBF1669.9 GB
Llama-4-Scout-17B-16E-Instruct108.6BBF1660.7 GB
Llama-3.2-90B-Vision-Instruct88.6BBF1649.5 GB
GLM-4.5V107.7BBF1660.2 GB

Other models

1 model.

ModelParametersPublished asVRAM needed
prism105.0BF8_E4M358.7 GB

How these numbers are computed

Required VRAM is the weight size at each precision times a flat 1.2 overhead, the same figure every model page shows. It does not include a long context: the KV cache grows with every token, so a model near the top of a tier can need the next one at long context. Read how much VRAM you need for LLMs for the method, the VRAM and quantization glossary entries for the terms, and the VRAM calculator to size a model that is not listed.

Looking for a model by job rather than by memory? Start with coding, reasoning, chat and assistants, vision-language or see the full models directory.

Other VRAM tiers

Start ComfyUI on a rented GPU.

© 2026 Aquanode. All rights reserved.

All trademarks, logos and brand names are the property of their respective owners.