NVIDIA A100 80GB (PCIe) for local LLMs
Written by Jakub Rusinowski · Last updated
With 80 GB at 1,935 GB/s, the A100 80GB (PCIe) runs 130 catalogued models at Q4_K_M with 8K context. The largest that fits is Devstral-2 123B (~75.1 GB), and the top pick is GPT-OSS 120B at 129–248 tok/s.
The datacenter workhorse of the 2021–2023 AI wave, now the value pick for on-prem business inference: 80 GB HBM2e at 1.9 TB/s runs Llama 3.3 70B at Q4 entirely on one card. No official MSRP — single-unit street pricing has fallen well below launch levels as H100/B200 fleets displace it, and the used/refurb market is active. PCIe version shown; SXM needs an HGX carrier board.
Models that run on the A100 80GB (PCIe)
Q4_K_M, 8K context, 80 GB usable. Ranked by quality and speed.
| Model | Memory · marker = 80 GB | VRAM | Speed |
|---|---|---|---|
| GPT-OSS 120B GPT-OSS | 71.9 GB | 129–248 tok/s | |
| Llama 4.5 Scout Llama 4.5 | 66.6 GB | 52–101 tok/s | |
| Llama 4 Scout 17B Llama 4 | 68.2 GB | 57–109 tok/s | |
| Llama 3.2 Vision 90B Llama 3.2 Vision | 57.8 GB | 14–27 tok/s | |
| Qwen3-Coder-Next (80B-A3B MoE) Qwen3-Coder | 49.1 GB | 118–226 tok/s | |
| Kolibri 1 Kolibri | 48.4 GB | 166–319 tok/s | |
| Llama 3.3 70B Instruct Llama 3.3 | 45.7 GB | 17–33 tok/s | |
| Cogito v1 70B Cogito v1 | 45.7 GB | 17–33 tok/s | |
| Apertus 70B Apertus | 43.1 GB | 17–33 tok/s | |
| Cosmos 3 Super Cosmos 3 | 39.4 GB | 34–65 tok/s | |
| K2 Horizon MoVA 36B-A4B K2 Horizon | 25 GB | 123–235 tok/s | |
| Qwen 3.6 35B-A3B Qwen 3.6 | 22.1 GB | 176–338 tok/s |
Buy it or rent the same memory
Buy the card, or rent a GPU with the same memory by the hour to try models first.
or compare on Vast.ai from $0.77/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
Speed vs other GPUs
Llama 3.1 8B, Q4_K_M. Estimated ranges. How this is calculated
Specifications
Specs last updated 2026-07-21.
- Memory
- 80 GB
- Memory bandwidth
- 1,935 GB/s
- Architecture
- Ampere GA100
- Series
- Server
- Board power
- 300 W
- Release year
- 2021
- Launch price
- $15,000
- Usable for models
- 80 GB
Similar GPUs
Frequently asked questions
Can the NVIDIA A100 80GB (PCIe) run local LLMs?
Yes. With 80 GB (80 GB usable by a model) it runs 130 of the catalogued models at Q4_K_M with 8K context; the largest is Devstral-2 123B, needing about 75.1 GB.
How fast is the NVIDIA A100 80GB (PCIe) for AI inference?
It is estimated to run Llama 3.1 8B at 102–196 tok/s at Q4_K_M. Llama 3.3 70B is estimated at 18–34 tok/s. These are modelled estimates from memory bandwidth, not measurements; the methodology page shows the formula.
What LLMs can I run on 80 GB?
Among the best that fit: GPT-OSS 120B, Llama 4.5 Scout, Llama 4 Scout 17B, Llama 3.2 Vision 90B, Qwen3-Coder-Next (80B-A3B MoE). The quickest start is Ollama: ollama run llama3.3:70b.
Keep going
Can I run it on the A100 80GB (PCIe)?
More for this card
Tools