Home › GPUs & Macs › NVIDIA RTX 5090 vs NVIDIA RTX PRO 6000 Blackwell Max-Q
NVIDIA RTX 5090 vs NVIDIA RTX PRO 6000 Blackwell Max-Q for local LLMs
Across 4 tests measured on both, the NVIDIA RTX 5090 is faster in 4 and generates a median 1.28× the tokens/s of the NVIDIA RTX PRO 6000 Blackwell Max-Q.
On gpt-oss 20B (MXFP4, llama-bench) the NVIDIA RTX 5090 generates 283 tokens/s and the NVIDIA RTX PRO 6000 Blackwell Max-Q 250 tokens/s (measured).
Same model, same quant, same test
| Model | Quant / test | RTX 5090 tok/s | RTX PRO 6000 Blackwell Max-Q tok/s | Faster | Sources and run details |
|---|---|---|---|---|---|
| gpt-oss 20B | MXFP4 llama-bench | 283 CUDA | 250 CUDA | RTX 5090 +13% | llama.cpp #15396 RTX 5090 · llama.cpp #15396 RTX PRO 6000 Blackwell Max-Q |
| Llama 3.1 8B Instruct | Q4_K_M LocalScore | 66.3 llamafile | 46.2 llamafile | RTX 5090 +44% | LocalScore RTX 5090 · LocalScore RTX PRO 6000 Blackwell Max-Q |
| Llama 3.2 1B Instruct | Q4_K_M LocalScore | 120 llamafile | 106 llamafile | RTX 5090 +13% | LocalScore RTX 5090 · LocalScore RTX PRO 6000 Blackwell Max-Q |
| Qwen2.5 14B Instruct | Q4_K_M LocalScore | 41.8 llamafile | 23.7 llamafile | RTX 5090 +76% | LocalScore RTX 5090 · LocalScore RTX PRO 6000 Blackwell Max-Q |
Generation speed in tokens/s, best measured backend for each chip (shown under the number). Backends can differ between the two chips; check them before drawing conclusions. Measured results only, nothing estimated.
Specs
| NVIDIA RTX 5090 | NVIDIA RTX PRO 6000 Blackwell Max-Q | |
|---|---|---|
| Memory | 32 GB | 96 GB |
| Memory bandwidth | 1792 GB/s | 1792 GB/s |
All results for the NVIDIA RTX 5090 · All results for the NVIDIA RTX PRO 6000 Blackwell Max-Q · More comparisons