What LLMs Can the A100 80GB Run?

The A100 80GB has 80 GB of VRAM and 2039 GB/s of bandwidth. It runs 50 of 58 models fully at Q4_K_M with 8K context, up to Qwen 3.5 122B-A10B, and generates about 227 tokens/s on an 8B model.

NVIDIA · Ampere · Released 2020-11 · Updated October 14, 2026

51 of 58 models fit

A100 80GB (80.0 GB) at Q4_K_M with 8K context. Largest that fits fully: Qwen 3.5 122B-A10B.

ModelNeedsStatusSpeedBest quant that fits
Qwen 3.5 122B-A10B70.2 GBFits~105 tok/sQ4_K_M
Mistral Small 4 119B68.2 GBFits~151 tok/sQ4_K_M
gpt-oss-120b67.2 GBFits~165 tok/sMXFP4 (official)
Llama 4 Scout64.0 GBFits~76 tok/sQ5_K_M
GLM-4.5-Air62.1 GBFits~94 tok/sQ5_K_M
Qwen3-Next 80B-A3B46.3 GBFits~159 tok/sQ6_K
Qwen 2.5 72B44.5 GBFits~29 tok/sQ6_K
Llama 3.3 70B43.2 GBFits~29 tok/sQ8_0
DeepSeek R1 Distill Llama 70B43.2 GBFits~29 tok/sQ8_0
Mixtral 8x7B28.0 GBFits~105 tok/sQ8_0
Qwen 2.5 32B21.4 GBFits~61 tok/sFP16 / BF16
Qwen 2.5 Coder 32B21.4 GBFits~61 tok/sFP16 / BF16
DeepSeek R1 Distill Qwen 32B21.4 GBFits~61 tok/sFP16 / BF16
Qwen 3 32B21.4 GBFits~61 tok/sFP16 / BF16
Qwen 3.5 35B-A3B21.0 GBFits~182 tok/sFP16 / BF16
Qwen 3.6 35B-A3B21.0 GBFits~182 tok/sFP16 / BF16
Gemma 4 31B19.8 GBFits~64 tok/sFP16 / BF16
Qwen 3 30B-A3B18.8 GBFits~151 tok/sFP16 / BF16
Qwen3-Coder 30B-A3B18.8 GBFits~151 tok/sFP16 / BF16
Gemma 3 27B17.6 GBFits~73 tok/sFP16 / BF16
Qwen 3.5 27B16.9 GBFits~75 tok/sFP16 / BF16
Qwen 3.6 27B16.9 GBFits~75 tok/sFP16 / BF16
Qwen 3.8 27B16.9 GBFits~75 tok/sFP16 / BF16
Mistral Small 3.1 24B15.6 GBFits~84 tok/sFP16 / BF16
Gemma 4 26B-A4B15.6 GBFits~202 tok/sFP16 / BF16
gpt-oss-20b12.9 GBFits~242 tok/sMXFP4 (official)
Qwen 2.5 14B10.7 GBFits~128 tok/sFP16 / BF16
DeepSeek R1 Distill Qwen 14B10.7 GBFits~128 tok/sFP16 / BF16
Phi-4 14B10.6 GBFits~129 tok/sFP16 / BF16
Qwen 3 14B10.5 GBFits~130 tok/sFP16 / BF16
Ministral 3 14B9.7 GBFits~141 tok/sFP16 / BF16
Mistral Nemo 12B9.0 GBFits~154 tok/sFP16 / BF16
Gemma 3 12B8.8 GBFits~153 tok/sFP16 / BF16
Ministral 3 8B6.7 GBFits~215 tok/sFP16 / BF16
Qwen 3 8B6.6 GBFits~219 tok/sFP16 / BF16
Qwen 3.5 9B6.4 GBFits~220 tok/sFP16 / BF16
Llama 3.1 8B6.3 GBFits~227 tok/sFP16 / BF16
DeepSeek R1 Distill Llama 8B6.3 GBFits~227 tok/sFP16 / BF16
Mistral 7B v0.35.7 GBFits~247 tok/sFP16 / BF16
Gemma 4 E4B5.7 GBFits~241 tok/sFP16 / BF16
Qwen 2.5 7B5.6 GBFits~254 tok/sFP16 / BF16
DeepSeek R1 Distill Qwen 7B5.6 GBFits~254 tok/sFP16 / BF16
Qwen 3 4B4.2 GBFits~377 tok/sFP16 / BF16
Phi-4 Mini4.1 GBFits~403 tok/sFP16 / BF16
Gemma 4 E2B4.0 GBFits~369 tok/sFP16 / BF16
Gemma 3 4B3.7 GBFits~406 tok/sFP16 / BF16
Qwen 3.5 4B3.7 GBFits~420 tok/sFP16 / BF16
Ministral 3 3B3.5 GBFits~463 tok/sFP16 / BF16
Llama 3.2 3B3.5 GBFits~476 tok/sFP16 / BF16
Llama 3.2 1B1.7 GBFits~1207 tok/sFP16 / BF16
Gemma 3 1B1.6 GBFits~1354 tok/sFP16 / BF16
Llama 3.1 405B234 GB32% on GPU~0.25 tok/s–
Qwen 3.5 397B-A17B226 GB33% on GPU~5.6 tok/s–
DeepSeek V4 Flash161 GB47% on GPU~9.3 tok/s–
Qwen 3 235B-A22B135 GB56% on GPU~6.1 tok/s–
MiniMax-M2133 GB57% on GPU~12 tok/s–
Kimi K2.6566 GBToo small––
DeepSeek R1 671B380 GBToo small––

A100 80GB for local AI

Still widely rented in the cloud. 80 GB holds 70B models at Q6_K or FP8-sized weights, with room for long contexts.

Among the 2 devices we track with about 80 GB of usable memory, the A100 80GB is the slowest at generating text, because generation speed follows memory bandwidth. Others in the same class: H100 80GB SXM. They all hold the same models; the faster ones simply answer sooner.

Our picks for the A100 80GB

The model we would run first for each job, chosen from our recommendations by what fits fully with an 8K context. "Best quant" is the highest quality that fits; "max context" is at Q4_K_M or the native format.

TaskModelBest quantMax contextSpeed
Best all-rounderQwen 3.5 122B-A10BQ4_K_M219K~105 tok/s
CodingGLM-4.5-AirQ5_K_M80K~94 tok/s
Fastest repliesQwen 3.5 122B-A10BQ4_K_M219K~105 tok/s
Long documents (64K+)Qwen 3.5 122B-A10BQ4_K_M219K~105 tok/s
Images in, text outGemma 4 31BFP16 / BF16256K~64 tok/s
Image generationFLUX.2 [dev]FP8––
SpecValue
Memory80 GB
Memory bandwidth2039 GB/s
ArchitectureAmpere
Released2020-11
Llama 3.1 8B Q4_K_M speed~227 tok/s
Largest model that fitsQwen 3.5 122B-A10B

Image and video models on the A100 80GB

Peak VRAM with model swapping (ComfyUI default on smaller cards).

ModelFP8FP8 fitGGUF Q4Q4 fit
Stable Diffusion 1.51.8 GBFits1.7 GBFits
SDXL 1.04.1 GBFits4.1 GBFits
Stable Diffusion 3.5 Medium6.4 GBFits6.4 GBFits
Stable Diffusion 3.5 Large8.6 GBFits6.4 GBFits
FLUX.1 [dev]12.1 GBFits7.3 GBFits
FLUX.1 [schnell]12.1 GBFits7.3 GBFits
Qwen-Image20.4 GBFits12.3 GBFits
Wan 2.2 A14B (T2V / I2V)18.8 GBFits13.2 GBFits
Wan 2.2 TI2V 5B8.0 GBFits6.5 GBFits
HunyuanVideo18.3 GBFits13.1 GBFits
FLUX.1 Kontext [dev]12.3 GBFits7.5 GBFits
FLUX.2 [dev]31.0 GBFits23.6 GBFits
Z-Image Turbo6.8 GBFits4.9 GBFits
Qwen-Image-Edit 250920.5 GBFits12.4 GBFits
LTX-219.2 GBFits12.6 GBFits

Upgrade path

The next step up is the NVIDIA H200 141GB with 141 GB and 135% more bandwidth.

Comparisons

A100 80GB for LLMs: common questions

Can the A100 80GB run Llama 3.3 70B?

Yes. Llama 3.3 70B needs about 43.2 GB at Q4_K_M with 8K context, and the best quality that fits is Q8_0.

Can the A100 80GB run Qwen 3 32B?

Yes. Qwen 3 32B needs about 21.4 GB at Q4_K_M with 8K context, and the best quality that fits is FP16 / BF16.

Can the A100 80GB run Qwen 3 14B?

Yes. Qwen 3 14B needs about 10.5 GB at Q4_K_M with 8K context, and the best quality that fits is FP16 / BF16.

How fast is the A100 80GB for LLMs?

About 227 tokens per second for an 8B model at Q4_K_M, based on its 2039 GB/s memory bandwidth. Larger models scale down roughly in proportion to their size.

Related