RTX PRO 4500 Blackwell
NVIDIA RTX PRO 4500 Blackwell: 32 GB of memory at 896 GB/s. Below: the open models it can run, estimated at 32 GB with an 8K context. 32 GB at 896 GB/s in a 200 W card: 27–32B models at higher-quality quantizations.
Last updated: 2026-10-16
Specifications
What it can run
Best-quality quantization that fits with a 8K context, and the estimated generation speed. Estimates, not benchmarks. See the methodology.
| Model | Params | Best quant | Memory | tok/s (est.) |
|---|---|---|---|---|
| Qwen3.5-35B-A3B | 35B (3B) | Q6_K | 28.9 GiB | 168 |
| Qwen3.6-35B-A3B | 35B (3B) | Q6_K | 28.9 GiB | 168 |
| Qwen3-32B | 32.8B | Q6_K | 28.9 GiB | 22 |
| DeepSeek-R1-Distill-Qwen-32B | 32.76B | Q6_K | 28.9 GiB | 22 |
| Gemma 4 31B IT | 30.7B | Q6_K | 26.3 GiB | 24 |
| Qwen3-30B-A3B (2507) | 30.5B (3.3B) | Q6_K | 25.8 GiB | 158 |
| Qwen3-Coder-30B-A3B-Instruct | 30.5B (3.3B) | Q6_K | 25.8 GiB | 158 |
| GLM-4.7-Flash (30B-A3B) | 30B (3B) | Q6_K | 25.1 GiB | 168 |
| Qwen3.5-27B | 27B | Q8_0 | 29.1 GiB | 21 |
| Qwen3.6-27B | 27B | Q8_0 | 29.1 GiB | 21 |
| Qwen3.8-27B | 27B | Q8_0 | 29.1 GiB | 21 |
| Gemma 3 27B IT | 27B | Q8_0 | 29.6 GiB | 21 |
| Gemma 4 26B A4B IT | 25.2B (3.8B) | Q8_0 | 27.0 GiB | 119 |
| Mistral Small 3.2 24B Instruct (2506) | 24B | Q8_0 | 26.7 GiB | 23 |
| Magistral Small 1.2 (2509) | 24B | Q8_0 | 26.7 GiB | 23 |
| Devstral Small 2 24B Instruct (2512) | 24B | Q8_0 | 26.7 GiB | 23 |
| gpt-oss-20b | 21B (3.6B) | Q8_0 | 22.5 GiB | 123 |
| Phi-4 (14B) | 14B | FP16 | 29.4 GiB | 21 |
| Gemma 3 12B IT | 12B | FP16 | 24.8 GiB | 25 |
| Gemma 4 12B Unified IT | 11.95B | FP16 | 24.2 GiB | 25 |
| Qwen3.5-9B | 9B | FP16 | 18.4 GiB | 33 |
| Qwen3-8B | 8.2B | FP16 | 17.7 GiB | 36 |
| DeepSeek-R1-0528-Qwen3-8B | 8.19B | FP16 | 17.6 GiB | 36 |
| Llama 3.1 8B Instruct | 8B | FP16 | 17.1 GiB | 36 |
| Gemma 4 E4B IT | 8B | FP16 | 16.4 GiB | 36 |
| Llama 3.2 3B Instruct | 3.21B | FP16 | 7.7 GiB | 82 |
| Kimi K2.6 | 1000B (32B) | Doesn't fit | – | – |
| DeepSeek-R1-0528 (671B) | 671B (37B) | Doesn't fit | – | – |
| DeepSeek-V4-Flash | 284B (13B) | Doesn't fit | – | – |
| Qwen3.8-Flash-Next | 180B (6B) | Doesn't fit | – | – |
| Qwen3.5-122B-A10B | 122B (10B) | Doesn't fit | – | – |
| Mistral Small 4 119B (2603) | 119B (6.5B) | Doesn't fit | – | – |
| gpt-oss-120b | 117B (5.1B) | Doesn't fit | – | – |
| Llama 4 Scout 17B-16E Instruct | 109B (17B) | Doesn't fit | – | – |
| GLM-4.5-Air | 106B (12B) | Doesn't fit | – | – |
| Qwen3-Coder-Next (80B-A3B) | 80B (3B) | Doesn't fit | – | – |
| Llama 3.3 70B Instruct | 70B | Doesn't fit | – | – |
Prices
Latest price at each shop, as listed on the day we checked.
| Region | Shop | Listing | Price | Per GB | Checked |
|---|---|---|---|---|---|
| Taiwan | 原價屋 CoolPC | Leadtek RTX PRO 4500 Blackwell 32GBLine on the shop's public price list (no per-product page); list stamped 2026/10/15 9:31 | NT$189,000 | NT$5,906/GB | 2026-10-15 |
- Taiwan value: about 102.7 tok/s on Llama 3.1 8B Instruct (Q4_K_M, 8K), or 0.5 tok/s per NT$1,000.
A price history chart appears once we have checked this device on two or more days.
All prices →Frequently asked questions
Which LLMs can the RTX PRO 4500 Blackwell run?
With 32 GB and an 8K context, the largest models that fit include: Qwen3.5-35B-A3B, Qwen3.6-35B-A3B, Qwen3-32B, DeepSeek-R1-Distill-Qwen-32B, Gemma 4 31B IT. See the full table below.
How fast is the RTX PRO 4500 Blackwell for local LLMs?
Generation speed is mostly set by memory bandwidth (896 GB/s). For Llama 3.1 8B Instruct at FP16, expect roughly 36 tokens/s; bigger models are slower.
Can the RTX PRO 4500 Blackwell run 70B models?
Llama 3.3 70B Instruct does not fit on the RTX PRO 4500 Blackwell (32 GB), even at Q4_K_M with an 8K context.