NVIDIA GeForce RTX 3080 Ti for local LLMs
Written by Jakub Rusinowski · Last updated
With 12 GB of GDDR6X at 912 GB/s, the RTX 3080 Ti runs 76 catalogued models at Q4_K_M with 8K context. The largest that fits is Qwen 3 14B (~11.1 GB), and the top pick is Cosmos 3 Nano at 51–99 tok/s.
12 GB at 912 GB/s — unusually high bandwidth for the capacity, which suits models that fit in 12 GB and are bandwidth-bound. Used-market only.
Models that run on the RTX 3080 Ti
Q4_K_M, 8K context, 12 GB usable. Ranked by quality and speed.
| Model | Memory · marker = 12 GB | VRAM | Speed |
|---|---|---|---|
| Cosmos 3 Nano Cosmos 3 | 10.5 GB | 51–99 tok/s | |
| Ministral 3 14B Ministral 3 | 9.3 GB | 35–67 tok/s | |
| DeepSeek R1 Distill Qwen 14B DeepSeek R1 | 10.6 GB | 35–67 tok/s | |
| Mellum2.1 12B Thinking Mellum2.1 | 9 GB | 134–257 tok/s | |
| Gemma 4 12B (Unified) Gemma 4 | 8.7 GB | 42–82 tok/s | |
| Mistral NeMo 12B Mistral Family | 9.4 GB | 39–76 tok/s | |
| Gemma 3 12B Instruct Gemma 3 | 9.1 GB | 41–78 tok/s | |
| Bielik PL 11B v3.0 Instruct Bielik | 7.4 GB | 42–81 tok/s | |
| Llama 3.2 Vision 11B Llama 3.2 Vision | 8.5 GB | 43–83 tok/s | |
| Llama 3.2 11B Vision Instruct Llama 3.2 Family | 8.5 GB | 43–83 tok/s | |
| Falcon 3 10B Instruct Falcon 3 | 8.4 GB | 44–84 tok/s | |
| Qwen 3.5 9B Qwen 3.5 | 6.6 GB | 55–106 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 $0.35/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 2026-09-19.
- Memory
- 12 GB GDDR6X
- Memory bandwidth
- 912 GB/s
- Memory bus
- 384-bit
- Architecture
- Ampere GA102
- Series
- RTX 30-series
- Board power
- 350 W
- Release year
- 2021
- Launch price
- $1,199
- Compute backends
- CUDA
- Usable for models
- 12 GB
Similar GPUs
Frequently asked questions
Can the NVIDIA GeForce RTX 3080 Ti run local LLMs?
Yes. With 12 GB (12 GB usable by a model) it runs 76 of the catalogued models at Q4_K_M with 8K context; the largest is Qwen 3 14B, needing about 11.1 GB.
How fast is the NVIDIA GeForce RTX 3080 Ti for AI inference?
It is estimated to run Llama 3.1 8B at 59–112 tok/s at Q4_K_M. Llama 3.3 70B does not fit: it needs about 44 GB against 12 GB usable. These are modelled estimates from memory bandwidth, not measurements; the methodology page shows the formula.
What LLMs can I run on 12 GB?
Among the best that fit: Cosmos 3 Nano, Ministral 3 14B, DeepSeek R1 Distill Qwen 14B, Mellum2.1 12B Thinking, Gemma 4 12B (Unified).