NVIDIA · 2022-12
Best local LLMs for the RTX 6000 Ada
48 GB VRAM · 960 GB/s · ~47 GB usable.
Check RTX 6000 Ada price →Try it with your context & use case
Preset to the RTX 6000 Ada. Change the context length or filter by use case.
27 models run well and 1 run tight on RTX 6000 Ada (any context — just fitting the weights) — 47 GB usable.
Sort:
| Model | Size | Fit | Best quant | Needs | Memory | Speed |
|---|---|---|---|---|---|---|
| Qwen2.5 72B | 72.7B | Runs well | Q4_K_M | 43 GB | ~16 tok/s | |
| Llama 3.3 70B | 70.6B | Runs well | Q4_K_M | 42 GB | ~16 tok/s | |
| DeepSeek-R1 Distill Llama 70B | 70.6B | Runs well | Q4_K_M | 42 GB | ~16 tok/s | |
| Mixtral 8x7B (MoE)MoE | 46.7B | Runs well | Q6_K | 38 GB | ~42 tok/s | |
| Qwen3 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~20 tok/s | |
| Qwen2.5 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~20 tok/s | |
| Qwen2.5-Coder 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~20 tok/s | |
| DeepSeek-R1 Distill Qwen 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~20 tok/s | |
| Qwen3 30B-A3B (MoE)MoE | 30.5B | Runs well | Q8_0 | 32 GB | ~128 tok/s | |
| Gemma 3 27B | 27.4B | Runs well | Q8_0 | 29 GB | ~24 tok/s | |
| Gemma 2 27B | 27.2B | Runs well | Q8_0 | 29 GB | ~24 tok/s | |
| Mistral Small 3 24B | 23.6B | Runs well | FP16 | 46 GB | ~15 tok/s | |
| Qwen3 14B | 14.8B | Runs well | FP16 | 29 GB | ~23 tok/s | |
| Qwen2.5 14B | 14.8B | Runs well | FP16 | 29 GB | ~23 tok/s | |
| DeepSeek-R1 Distill Qwen 14B | 14.8B | Runs well | FP16 | 29 GB | ~23 tok/s | |
| Phi-4 14B | 14.7B | Runs well | FP16 | 29 GB | ~24 tok/s | |
| Mistral Nemo 12B | 12.2B | Runs well | FP16 | 24 GB | ~28 tok/s | |
| Gemma 3 12B | 12.2B | Runs well | FP16 | 24 GB | ~28 tok/s | |
| Gemma 2 9B | 9.2B | Runs well | FP16 | 19 GB | ~37 tok/s | |
| Qwen3 8B | 8.2B | Runs well | FP16 | 17 GB | ~42 tok/s | |
| Llama 3.1 8B | 8.0B | Runs well | FP16 | 16 GB | ~43 tok/s | |
| Qwen2.5-Coder 7B | 7.6B | Runs well | FP16 | 15 GB | ~45 tok/s | |
| Mistral 7B v0.3 | 7.3B | Runs well | FP16 | 15 GB | ~48 tok/s | |
| Gemma 3 4B | 4.3B | Runs well | FP16 | 9.1 GB | ~80 tok/s | |
| Qwen3 4B | 4B | Runs well | FP16 | 8.5 GB | ~86 tok/s | |
| Llama 3.2 3B | 3.2B | Runs well | FP16 | 7.0 GB | ~108 tok/s | |
| Llama 3.2 1B | 1.2B | Runs well | FP16 | 3.2 GB | ~279 tok/s | |
| Llama 4 Scout (MoE)MoE | 109B | Runs (tight) | Q2_K | 45 GB | ~63 tok/s | |
| DeepSeek-R1 671B-A37B (MoE)MoE | 671B | Won't fit | — | 273 GB | — | |
| Qwen3 235B-A22B (MoE)MoE | 235B | Won't fit | — | 96 GB | — |
Estimates, computed in your browser — VRAM, quantisation and speed vary with your runtime and settings. How we estimate →
Top picks for the RTX 6000 Ada
Every model on the RTX 6000 Ada
| Model | Size | Fit | Best quant | Needs | Speed |
|---|---|---|---|---|---|
| Qwen2.5 72B | 72.7B | Runs well | Q4 | 46 GB | ~16 tok/s |
| Llama 3.3 70B | 70.6B | Runs well | Q4 | 45 GB | ~16 tok/s |
| DeepSeek-R1 Distill Llama 70B | 70.6B | Runs well | Q4 | 45 GB | ~16 tok/s |
| Mixtral 8x7B (MoE)MoE | 46.7B | Runs well | Q6 | 39 GB | ~42 tok/s |
| Qwen3 32B | 32.8B | Runs well | Q8 | 37 GB | ~20 tok/s |
| Qwen2.5 32B | 32.8B | Runs well | Q8 | 37 GB | ~20 tok/s |
| Qwen2.5-Coder 32B | 32.8B | Runs well | Q8 | 37 GB | ~20 tok/s |
| DeepSeek-R1 Distill Qwen 32B | 32.8B | Runs well | Q8 | 37 GB | ~20 tok/s |
| Qwen3 30B-A3B (MoE)MoE | 30.5B | Runs well | Q8 | 33 GB | ~128 tok/s |
| Gemma 3 27B | 27.4B | Runs well | Q8 | 33 GB | ~24 tok/s |
| Gemma 2 27B | 27.2B | Runs well | Q8 | 32 GB | ~24 tok/s |
| Mistral Small 3 24B | 23.6B | Runs well | Q8 | 26 GB | ~28 tok/s |
| Qwen3 14B | 14.8B | Runs well | FP16 | 31 GB | ~23 tok/s |
| Qwen2.5 14B | 14.8B | Runs well | FP16 | 31 GB | ~23 tok/s |
| DeepSeek-R1 Distill Qwen 14B | 14.8B | Runs well | FP16 | 31 GB | ~23 tok/s |
| Phi-4 14B | 14.7B | Runs well | FP16 | 31 GB | ~24 tok/s |
| Mistral Nemo 12B | 12.2B | Runs well | FP16 | 26 GB | ~28 tok/s |
| Gemma 3 12B | 12.2B | Runs well | FP16 | 27 GB | ~28 tok/s |
| Gemma 2 9B | 9.2B | Runs well | FP16 | 21 GB | ~37 tok/s |
| Qwen3 8B | 8.2B | Runs well | FP16 | 18 GB | ~42 tok/s |
| Llama 3.1 8B | 8.0B | Runs well | FP16 | 17 GB | ~43 tok/s |
| Qwen2.5-Coder 7B | 7.6B | Runs well | FP16 | 16 GB | ~45 tok/s |
| Mistral 7B v0.3 | 7.3B | Runs well | FP16 | 16 GB | ~48 tok/s |
| Gemma 3 4B | 4.3B | Runs well | FP16 | 10 GB | ~80 tok/s |
| Qwen3 4B | 4B | Runs well | FP16 | 9.6 GB | ~86 tok/s |
| Llama 3.2 3B | 3.2B | Runs well | FP16 | 7.8 GB | ~108 tok/s |
| Llama 3.2 1B | 1.2B | Runs well | FP16 | 3.4 GB | ~279 tok/s |
| Llama 4 Scout (MoE)MoE | 109B | Runs (tight) | Q2 | 46 GB | ~63 tok/s |
| DeepSeek-R1 671B-A37B (MoE)MoE | 671B | Won't fit | — | 303 GB | — |
| Qwen3 235B-A22B (MoE)MoE | 235B | Won't fit | — | 98 GB | — |
Similar hardware
FAQ
What is the best LLM for the RTX 6000 Ada?
For general use, Qwen2.5 72B is the strongest model that runs well on the RTX 6000 Ada. See the picks-by-use-case below for coding, reasoning and more.
How much can the RTX 6000 Ada run?
The RTX 6000 Ada has 48 GB of VRAM, of which about 47 GB is usable for a model. That runs 27 of our tracked models well and 1 more at a tight quantisation.
Estimates at 8K context — see how we compute these.