GeForce RTX 3080 (10 GB)
10 GB GDDR6X · 760 GB/s · 320 W
RTX 3080 10 GB · 64 GB RAM → Mistral Nemo 12B Instruct
GPU · NVIDIA
RTX 3080 for local LLMs: which models run?
With 10 GB of VRAM and 760 GB/s of bandwidth, the RTX 3080 runs models up to about Mistral Nemo 12B Instruct entirely on the GPU.
- VRAM10 GB GDDR6X
- Bandwidth760 GB/s
- Board power320 W
- ArchitectureAmpere
Specifications
- VRAM
- 10 GB
- Memory type
- GDDR6X
- Memory bus
- 320-bit
- Bandwidth
- 760 GB/s
- Board power
- 320 W
- Backends
- CUDA, Vulkan
- Interface
- PCIe 4.0 x16
- Released
- September 2020
- Segment
- Consumer
- FP16 compute
- Not available
Official specData updated 24 days agoSource: NVIDIA specifications
What the RTX 3080 runs
| Qwen2.5 7B InstructQ6_K · 8K~7.4 GB · 79–109 tok/s estimate | Excellent |
|---|---|
| Gemma 3 4B ITQ8_0 · 8K~5.4 GB · 107–156 tok/s estimate | Excellent |
| Qwen3 4BQ8_0 · 8K~6.2 GB · 95–138 tok/s estimate | Excellent |
| Phi-4 Mini Instruct 3.8BQ8_0 · 8K~6 GB · 102–147 tok/s estimate | Excellent |
| Llama 3.2 3B InstructQ8_0 · 8K~5.1 GB · 120–174 tok/s estimate | Excellent |
| Qwen3 1.7BQ8_0 · 8K~3.7 GB · 179–276 tok/s estimate | Excellent |
| Llama 3.2 1B InstructQ8_0 · 8K~2.5 GB · 291–443 tok/s estimate | Excellent |
| Qwen3 0.6BQ8_0 · 8K~2.5 GB · 280–491 tok/s estimate | Excellent |
| Mistral Nemo 12B InstructIQ4_XS · 8K~8.7 GB · 67–94 tok/s estimate | Compatible |
| Gemma 3 12B ITQ4_K_M · 8K~9.1 GB · 62–89 tok/s estimate | Compatible |
| Gemma 2 9B ITQ4_K_M · 8K~8.7 GB · 71–101 tok/s estimate | Compatible |
| Qwen3 8BQ6_K · 8K~8.6 GB · 68–95 tok/s estimate | Compatible |
| Llama 3.1 8B InstructQ6_K · 8K~8.3 GB · 71–99 tok/s estimate | Compatible |
| DeepSeek R1 Distill Llama 8BQ6_K · 8K~8.3 GB · 71–99 tok/s estimate | Compatible |
| Mistral 7B Instruct v0.3Q8_0 · 8K~9.2 GB · 63–86 tok/s estimate | Compatible |
| Qwen3 14BQ3_K_M · 8K~9.3 GB · 62–87 tok/s estimate | Compatible with limits |
| DeepSeek R1 Distill Qwen 14BQ3_K_M · 8K~9.6 GB · 60–85 tok/s estimate | Compatible with limits |
| Qwen2.5 14B InstructQ3_K_M · 8K~9.6 GB · 60–85 tok/s estimate | Compatible with limits |
| Phi-4 14BQ3_K_M · 8K~9.6 GB · 61–85 tok/s estimate | Compatible with limits |
| Qwen3 30B-A3BQ3_K_M · 8K~15.4 GB · 22–45 tok/s estimate | CPU/GPU offload |
| Qwen3 Coder 30B-A3B InstructQ3_K_M · 8K~16 GB · 21–41 tok/s estimate | CPU/GPU offload |
| Gemma 3 27B ITQ3_K_M · 8K~15.3 GB · 3.5–7 tok/s estimate | CPU/GPU offload |
| Gemma 2 27B ITQ3_K_M · 8K~16.5 GB · 3–6 tok/s estimate | CPU/GPU offload |
| Mistral Small 3.2 24B InstructQ3_K_M · 8K~13.3 GB · 5–10 tok/s estimate | CPU/GPU offload |
| gpt-oss-20bMXFP4 · 8K~13 GB · 28–55 tok/s estimate | CPU/GPU offload |
| Llama 4 Scout 17B-16E InstructQ3_K_M · 8K~55.7 GB · 2.5–5 tok/s estimate | Possible but slow |
| Mixtral 8x7B InstructQ3_K_M · 8K~21.6 GB · 5.5–11 tok/s estimate | Possible but slow |
| Qwen3 32BQ3_K_M · 8K~18.6 GB · 2.5–4.5 tok/s estimate | Possible but slow |
| DeepSeek R1 Distill Qwen 32BQ3_K_M · 8K~18.5 GB · 2.5–4.5 tok/s estimate | Possible but slow |
| Qwen2.5 32B InstructQ3_K_M · 8K~18.5 GB · 2.5–4.5 tok/s estimate | Possible but slow |
| Qwen2.5 Coder 32B InstructQ3_K_M · 8K~18.5 GB · 2.5–4.5 tok/s estimate | Possible but slow |
| QwQ 32BQ3_K_M · 8K~18.5 GB · 2.5–4.5 tok/s estimate | Possible but slow |
| Qwen2.5 72B InstructQ3_K_M · 8K~38.2 GB · 0.7–1.4 tok/s estimate | Not recommended |
| Llama 3.3 70B InstructQ3_K_M · 8K~37 GB · 0.7–1.4 tok/s estimate | Not recommended |
| DeepSeek R1 Distill Llama 70BQ3_K_M · 8K~37 GB · 0.7–1.4 tok/s estimate | Not recommended |
| DeepSeek R1 (671B)Q8_0 · 8K~699 GB · No data | Incompatible |
| Qwen3 235B-A22BQ8_0 · 8K~247 GB · No data | Incompatible |
| gpt-oss-120bMXFP4 · 8K~63.2 GB · No data | Incompatible |
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
10 GB of VRAM holds the weights and KV cache of models up to about Mistral Nemo 12B Instruct 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.
Where to buy a RTX 3080
Prices and links follow the market you choose.
No current offer imported for this market: the link opens a marketplace search.
Frequently asked questions
Which LLMs can the RTX 3080 run?
Entirely on the GPU: Qwen3 14B (Q3_K_M), DeepSeek R1 Distill Qwen 14B (Q3_K_M), Qwen2.5 14B Instruct (Q3_K_M), Phi-4 14B (Q3_K_M), Mistral Nemo 12B Instruct (IQ4_XS), Gemma 3 12B IT (Q4_K_M), Gemma 2 9B IT (Q4_K_M), Qwen3 8B (Q6_K). Larger models need CPU/GPU offload.
Can the RTX 3080 run a 70B model?
Not recommended: a 70B model at Q4_K_M needs about 45.1 GB; the RTX 3080 has 10 GB.
How is the speed estimated?
Token generation reads every active weight once per token, so it is limited by memory bandwidth (760 GB/s here). We apply an efficiency range per backend and label the result as an estimate until real benchmarks exist.