# NVIDIA RTX PRO 5000 Blackwell 72GB for local LLMs

> 72GB on one 300W card fits gpt-oss-120b whole, for a little over half the price of an RTX PRO 6000.

The RTX PRO 5000 72GB has 72GB of GDDR7 ECC and 1,344 GB/s of memory bandwidth, which decides how big a model fits and how fast it generates text. It comfortably runs models up to about 122B parameters: the largest in our catalog is Qwen3.5 122B-A10B at Q3_K_M (62GB with 8K context, ~99 tokens/sec). Our top all-round pick for it is Qwen3.8 27B (Q8_0, ~35 tokens/sec).

## Key specs

- Memory: 72GB GDDR7 ECC
- Memory bandwidth: 1,344 GB/s
- Architecture: Blackwell (2025)
- Power: 300W
- Typical price: ~$9,100 (checked 2026-10)
- Software: CUDA, Vulkan

## Best models for this GPU

- **Best overall (also coding, reasoning, vision):** Qwen3.8 27B at Q8_0, ~31.6GB with 8K context, ~35 tokens/sec
- **Runner-up:** Qwen3.6 35B-A3B at Q8_0, ~41.0GB with 8K context, ~130 tokens/sec
- **Fastest good model:** gpt-oss-20b at MXFP4, ~15.0GB with 8K context, ~160 tokens/sec
- **Largest model that fits:** Qwen3.5 122B-A10B at Q3_K_M, ~62.1GB with 8K context, ~99 tokens/sec

## Compatibility (best quantization that fits, 8K context)

| Model | Params | Quant | Memory | Speed | Verdict |
| --- | --- | --- | --- | --- | --- |
| Gemma 4 E4B | 8B | Q8_0 | 10.5GB | ~100 tok/s | Runs well |
| Qwen3.5 9B | 9B | Q8_0 | 12.0GB | ~90 tok/s | Runs well |
| Phi-4 | 14.7B | Q8_0 | 18.3GB | ~65 tok/s | Runs well |
| gpt-oss-20b | 20.9B (3.6B active) | MXFP4 | 15.0GB | ~160 tok/s | Runs well |
| Gemma 4 26B A4B | 26B (3.8B active) | Q8_0 | 29.5GB | ~115 tok/s | Runs well |
| Qwen3.8 27B | 27B | Q8_0 | 31.6GB | ~35 tok/s | Runs well |
| Gemma 4 31B | 31B | Q8_0 | 35.9GB | ~32 tok/s | Runs well |
| Qwen3.6 35B-A3B | 35B (3B active) | Q8_0 | 41.0GB | ~130 tok/s | Runs well |
| GLM-4.5-Air | 106B (12B active) | Q4_K_M | 68.3GB | ~73 tok/s | Fits, tight |
| Llama 4 Scout | 109B (17B active) | Q3_K_M | 56.3GB | ~66 tok/s | Runs well |
| gpt-oss-120b | 117B (5.1B active) | MXFP4 | 68.3GB | ~145 tok/s | Fits, tight |
| Qwen3.5 122B-A10B | 122B (10B active) | Q3_K_M | 62.1GB | ~99 tok/s | Runs well |
| DeepSeek V4 Flash | 284B (13B active) | - | - | - | Won't fit |
| GLM-5.3-Flash | 320B (18B active) | - | - | - | Won't fit |
| GLM-5.3 | 744B (40B active) | - | - | - | Won't fit |

## Pros
- 72GB on a single card
- 1,344 GB/s GDDR7 bandwidth
- 300W design suits multi-GPU towers
- ECC memory and professional drivers

## Cons
- Costs $9,100 or more
- Less memory than the RTX PRO 6000 for the largest models

Full page: https://llamabuilds.ai/gpus/rtx-pro-5000-blackwell-72gb · Best models: https://llamabuilds.ai/gpus/rtx-pro-5000-blackwell-72gb/best-models

Sources: [NVIDIA blog: RTX PRO 5000 72GB Blackwell now generally available](https://blogs.nvidia.com/blog/rtx-pro-5000-72gb-blackwell-gpu/), [RunAIHome RTX PRO 5000 72GB pricing and fit analysis](https://runaihome.com/blog/rtx-pro-5000-72gb-local-ai-2026/)

---
Source: https://llamabuilds.ai/gpus/rtx-pro-5000-blackwell-72gb · Updated October 14, 2026 · Llama Builds. Speeds marked ~ are estimates from our bandwidth model (https://llamabuilds.ai/methodology); measured results are labeled as such.