RTX 5090 vs RTX PRO 6000 Blackwell x4
Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.
RTX 509023 sourced runs · best 446.7 tok/s
RTX PRO 6000 Blackwell x49 verified runs · best 231.6 tok/s
Specs and evidence
| RTX 5090 | RTX PRO 6000 Blackwell x4 | |
|---|---|---|
| VRAM | 32 GB | 384 GB |
| Sourced runs on the board | 23 | 9 |
| Models measured | 16 | 1 |
| Best output tok/s | 446.7 tok/s | 231.6 tok/s |
| Highest measured peak VRAM | 31.3 GB | — |
| Street price (latest snapshot) | — | — |
The measured matrix
| Model | RTX 5090 | RTX PRO 6000 Blackwell x4 |
|---|---|---|
| NVIDIA-Nemotron-3.5-Lightning-30B-A3B | 446.7 | — |
| Qwen3.8-27B | 435.1 | — |
| Ornith-1.0-35B-PrismaAURA-4.75bit-vllm-MTP | 283.1 | — |
| Kimi-Linear-48B-A3B-Instruct | 280.8 | — |
| Ornith-1.5-35B-A3B | 258.8 | — |
| North-Mini-Code-1.0 | 258.4 | — |
| DeepSeek-V4.1-Flash | — | 231.6 |
| Qwen3.6-35B-A3B-MTP | 222.9 | — |
| Laguna-XS-2.1 | 216.7 | — |
| KAT-Coder-V2.5-Dev-AWQ-W4A16-ASYM | 201.2 | — |
| Ornith-1.0-35B | 196.6 | — |
| Muse-Glimmer-30B | 124.8 | — |
| Qwen3.6-27B | 120.9 | — |
| Huihui-Qwen3.8-27B-abliterated | 117.7 | — |
| Qwen3.8-27B-5090-goldilocks | 93.5 | — |
| Qwen3.6-27B-MTP | 28.3 | — |
| DeepSeek-V4-Flash-0731 | 26.4 | — |