# RTX 6000 Ada Generation

> OnPrem World · https://onpremworld.com/en/hardware/nvidia-rtx-6000-ada · Last updated: 2026-10-14 · CC BY 4.0

The same 48 GB as the RTX A6000 with 25% more bandwidth (960 GB/s), so 70B models at 4-bit generate proportionally faster.

## Specifications

|  |  |
|---|---|
| Memory | 48 GB GDDR6 (ECC) |
| Memory bandwidth | 960 GB/s |
| Power | 300 W |
| Launch | 2022-12 |
| Launch price | US$6,800 |
| Architecture | Ada Lovelace (AD102) |
| Software | CUDA |

## What it can run (48 GB, 8K)

Estimates, not benchmarks. See the methodology.

| Model | Parameters | Best quant | Total | tok/s (est.) |
|---|---|---|---|---|
| Llama 3.3 70B Instruct | 70B | Q4_K_M | 44.5 GiB | 16 |
| Qwen3.5-35B-A3B | 35B | Q8_0 | 37.0 GiB | 211 |
| Qwen3.6-35B-A3B | 35B | Q8_0 | 37.0 GiB | 211 |
| Qwen3-32B | 32.8B | Q8_0 | 36.6 GiB | 19 |
| DeepSeek-R1-Distill-Qwen-32B | 32.76B | Q8_0 | 36.5 GiB | 19 |
| Gemma 4 31B IT | 30.7B | Q8_0 | 33.5 GiB | 21 |
| Qwen3-30B-A3B (2507) | 30.5B | Q8_0 | 32.9 GiB | 192 |
| Qwen3-Coder-30B-A3B-Instruct | 30.5B | Q8_0 | 32.9 GiB | 192 |
| GLM-4.7-Flash (30B-A3B) | 30B | Q8_0 | 32.1 GiB | 211 |
| Qwen3.5-27B | 27B | Q8_0 | 29.1 GiB | 23 |
| Qwen3.6-27B | 27B | Q8_0 | 29.1 GiB | 23 |
| Qwen3.8-27B | 27B | Q8_0 | 29.1 GiB | 23 |
| Gemma 3 27B IT | 27B | Q8_0 | 29.6 GiB | 23 |
| Gemma 4 26B A4B IT | 25.2B | Q8_0 | 27.0 GiB | 166 |
| Mistral Small 3.2 24B Instruct (2506) | 24B | Q8_0 | 26.7 GiB | 26 |
| Magistral Small 1.2 (2509) | 24B | Q8_0 | 26.7 GiB | 26 |
| Devstral Small 2 24B Instruct (2512) | 24B | Q8_0 | 26.7 GiB | 26 |
| gpt-oss-20b | 21B | FP16 | 41.8 GiB | 93 |
| Phi-4 (14B) | 14B | FP16 | 29.4 GiB | 24 |
| Gemma 3 12B IT | 12B | FP16 | 24.8 GiB | 28 |
| Gemma 4 12B Unified IT | 11.95B | FP16 | 24.2 GiB | 28 |
| Qwen3.5-9B | 9B | FP16 | 18.4 GiB | 37 |
| Qwen3-8B | 8.2B | FP16 | 17.7 GiB | 41 |
| DeepSeek-R1-0528-Qwen3-8B | 8.19B | FP16 | 17.6 GiB | 41 |
| Llama 3.1 8B Instruct | 8B | FP16 | 17.1 GiB | 42 |
| Gemma 4 E4B IT | 8B | FP16 | 16.4 GiB | 42 |
| Llama 3.2 3B Instruct | 3.21B | FP16 | 7.7 GiB | 105 |
| Kimi K2.6 | 1000B | Doesn't fit | – | – |
| DeepSeek-R1-0528 (671B) | 671B | Doesn't fit | – | – |
| DeepSeek-V4-Flash | 284B | Doesn't fit | – | – |
| Qwen3.8-Flash-Next | 180B | Doesn't fit | – | – |
| Qwen3.5-122B-A10B | 122B | Doesn't fit | – | – |
| Mistral Small 4 119B (2603) | 119B | Doesn't fit | – | – |
| gpt-oss-120b | 117B | Doesn't fit | – | – |
| Llama 4 Scout 17B-16E Instruct | 109B | Doesn't fit | – | – |
| GLM-4.5-Air | 106B | Doesn't fit | – | – |
| Qwen3-Coder-Next (80B-A3B) | 80B | Doesn't fit | – | – |

## Prices

No prices recorded for this device yet.

## Frequently asked questions

### Which LLMs can the RTX 6000 Ada Generation run?

With 48 GB and an 8K context, the largest models that fit include: Llama 3.3 70B Instruct, Qwen3.5-35B-A3B, Qwen3.6-35B-A3B, Qwen3-32B, DeepSeek-R1-Distill-Qwen-32B. See the full table below.

### How fast is the RTX 6000 Ada Generation for local LLMs?

Generation speed is mostly set by memory bandwidth (960 GB/s). For Llama 3.1 8B Instruct at FP16, expect roughly 42 tokens/s; bigger models are slower.

### Can the RTX 6000 Ada Generation run 70B models?

Yes. Llama 3.3 70B Instruct fits on the RTX 6000 Ada Generation (48 GB) at Q4_K_M with an 8K context, using about 44.5 GiB, at an estimated 16 tokens/s.

## Sources

- https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/rtx-6000/proviz-print-rtx6000-datasheet-web-2504660.pdf
- https://www.nvidia.com/en-us/design-visualization/rtx-6000/
- https://nvidianews.nvidia.com/news/nvidias-new-ada-lovelace-rtx-gpu-arrives-for-designers-and-creators
- https://www.computerbase.de/2023-01/nvidia-a6000-ada-generation-6800-us-dollar/