Skip to content
Run LLM Locally

GPU · NVIDIA

A100 80GB for local LLMs: which models run?

With 80 GB of VRAM and 1935 GB/s of bandwidth, the A100 80GB runs models up to about gpt-oss-120b entirely on the GPU.

  • VRAM80 GB HBM2e
  • Bandwidth1,935 GB/s
  • Board power300 W
  • ArchitectureAmpere
ShareFacebookXWhatsAppLinkedInEmail

Specifications

VRAM
80 GB
Memory type
HBM2e
Memory bus
5120-bit
Bandwidth
1,935 GB/s
Board power
300 W
Backends
CUDA
Interface
PCIe 4.0 x16
Released
June 2021
Segment
Data center
FP16 compute
Not available

Official specData updated 24 days agoSource: NVIDIA specifications

What the A100 80GB runs

Popular models on A100 80GB + 64 GB RAM · 8K context
gpt-oss-120bMXFP4 · 8K~63.2 GB · 170–472 tok/s estimateExcellent
Qwen2.5 72B InstructQ6_K · 8K~62.1 GB · 22–30 tok/s estimateExcellent
Llama 3.3 70B InstructQ6_K · 8K~60.1 GB · 23–31 tok/s estimateExcellent
DeepSeek R1 Distill Llama 70BQ6_K · 8K~60.1 GB · 23–31 tok/s estimateExcellent
Mixtral 8x7B InstructQ8_0 · 8K~50.2 GB · 70–99 tok/s estimateExcellent
Qwen3 32BQ8_0 · 8K~37 GB · 36–50 tok/s estimateExcellent
DeepSeek R1 Distill Qwen 32BQ8_0 · 8K~37 GB · 36–50 tok/s estimateExcellent
Qwen2.5 32B InstructQ8_0 · 8K~37 GB · 36–50 tok/s estimateExcellent
Qwen2.5 Coder 32B InstructQ8_0 · 8K~37 GB · 36–50 tok/s estimateExcellent
QwQ 32BQ8_0 · 8K~37 GB · 36–50 tok/s estimateExcellent
Qwen3 30B-A3BQ8_0 · 8K~33.4 GB · 134–231 tok/s estimateExcellent
Qwen3 Coder 30B-A3B InstructQ8_0 · 8K~33.4 GB · 171–476 tok/s estimateExcellent
Gemma 3 27B ITQ8_0 · 8K~30.2 GB · 43–60 tok/s estimateExcellent
Gemma 2 27B ITQ8_0 · 8K~31.6 GB · 44–60 tok/s estimateExcellent
Mistral Small 3.2 24B InstructQ8_0 · 8K~26.6 GB · 51–70 tok/s estimateExcellent
gpt-oss-20bMXFP4 · 8K~13 GB · 244–672 tok/s estimateExcellent
Qwen3 14BQ8_0 · 8K~17.5 GB · 75–107 tok/s estimateExcellent
DeepSeek R1 Distill Qwen 14BQ8_0 · 8K~17.8 GB · 72–104 tok/s estimateExcellent
Qwen2.5 14B InstructQ8_0 · 8K~17.8 GB · 72–104 tok/s estimateExcellent
Phi-4 14BQ8_0 · 8K~17.6 GB · 75–106 tok/s estimateExcellent
Mistral Nemo 12B InstructQ8_0 · 8K~14.9 GB · 87–126 tok/s estimateExcellent
Gemma 3 12B ITQ8_0 · 8K~14.1 GB · 87–129 tok/s estimateExcellent
Gemma 2 9B ITQ8_0 · 8K~12.7 GB · 102–153 tok/s estimateExcellent
Qwen3 8BQ8_0 · 8K~10.5 GB · 119–179 tok/s estimateExcellent
Llama 3.1 8B InstructQ8_0 · 8K~10.2 GB · 125–186 tok/s estimateExcellent
DeepSeek R1 Distill Llama 8BQ8_0 · 8K~10.2 GB · 125–186 tok/s estimateExcellent
Qwen2.5 7B InstructQ8_0 · 8K~9.2 GB · 138–203 tok/s estimateExcellent
Mistral 7B Instruct v0.3Q8_0 · 8K~9.2 GB · 135–202 tok/s estimateExcellent
Gemma 3 4B ITQ8_0 · 8K~5.4 GB · 204–342 tok/s estimateExcellent
Qwen3 4BQ8_0 · 8K~6.2 GB · 184–305 tok/s estimateExcellent
Phi-4 Mini Instruct 3.8BQ8_0 · 8K~6 GB · 200–327 tok/s estimateExcellent
Llama 3.2 3B InstructQ8_0 · 8K~5.1 GB · 234–385 tok/s estimateExcellent
Qwen3 1.7BQ8_0 · 8K~3.7 GB · 311–566 tok/s estimateExcellent
Llama 3.2 1B InstructQ8_0 · 8K~2.5 GB · 517–925 tok/s estimateExcellent
Qwen3 0.6BQ8_0 · 8K~2.5 GB · 413–877 tok/s estimateExcellent
Llama 4 Scout 17B-16E InstructQ4_K_M · 8K~68.6 GB · 75–112 tok/s estimateCompatible
Qwen3 235B-A22BQ3_K_M · 8K~112 GB · 5–10 tok/s estimateCPU/GPU offload
DeepSeek R1 (671B)Q8_0 · 8K~699 GB · No dataIncompatible

For each model, the highest-precision quantization that runs entirely in fast memory; otherwise the one with the best result. Speeds are ranges; "estimate" unless a benchmark exists for this exact hardware.

EstimateData updated 24 days agoSource: Run LLM Locally engine

Limits in plain words

80 GB of VRAM holds the weights and KV cache of models up to about gpt-oss-120b at Q4_K_M. Bigger models spill into system RAM and slow down sharply: generation speed is then limited by RAM bandwidth, not by the GPU.

Frequently asked questions

Which LLMs can the A100 80GB run?
Entirely on the GPU: gpt-oss-120b (MXFP4), Llama 4 Scout 17B-16E Instruct (Q4_K_M), Qwen2.5 72B Instruct (Q6_K), Llama 3.3 70B Instruct (Q6_K), DeepSeek R1 Distill Llama 70B (Q6_K), Mixtral 8x7B Instruct (Q8_0), Qwen3 32B (Q8_0), DeepSeek R1 Distill Qwen 32B (Q8_0). Larger models need CPU/GPU offload.
Can the A100 80GB run a 70B model?
Excellent: a 70B model at Q4_K_M needs about 45.1 GB; the A100 80GB has 80 GB.
How is the speed estimated?
Token generation reads every active weight once per token, so it is limited by memory bandwidth (1935 GB/s here). We apply an efficiency range per backend and label the result as an estimate until real benchmarks exist.