RTX PRO 6000 Blackwell vs RX 7900 XTX
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 PRO 6000 Blackwell29 sourced runs · best 1271.3 tok/s
RX 7900 XTX6 verified runs · best 550.2 tok/s
Specs and evidence
| RTX PRO 6000 Blackwell | RX 7900 XTX | |
|---|---|---|
| VRAM | 96 GB | 24 GB |
| Sourced runs on the board | 29 | 6 |
| Models measured | 23 | 5 |
| Best output tok/s | 1271.3 tok/s | 550.2 tok/s |
| Highest measured peak VRAM | 72.7 GB | 22.9 GB |
| Street price (latest snapshot) | — | — |
Head-to-head
On the 2 models measured on both cards, RTX PRO 6000 Blackwell leads on 2, RX 7900 XTX on 0. The widest gap is on Qwen3.6-35B-A3B: RTX PRO 6000 Blackwell 330.2 tok/s vs RX 7900 XTX 150.0 tok/s (RTX PRO 6000 Blackwell 120% faster).
The measured matrix
| Model | RTX PRO 6000 Blackwell | RX 7900 XTX |
|---|---|---|
| MobileLLM-125M | 1271.3 | — |
| Qwen2.5-0.5B-Instruct | 1060.0 | — |
| MobileLLM-350M | 980.8 | — |
| SmolLM2-360M-Instruct | 968.1 | — |
| Qwen3-0.6B | 956.4 | — |
| Llama-3.2-1B-Instruct | 699.9 | — |
| gemma-3-1b-it | 664.3 | — |
| Qwen3.5-0.8B | — | 550.2 |
| Llama-3.2-3B-Instruct | 489.2 | — |
| LFM2.5-350M | — | 398.4 |
| Ling-3.0-tiny | 369.1 | — |
| Qwen3.6-35B-A3B | 330.2 | 150.0 |
| Ornith-1.0-35B | 309.7 | — |
| Qwen3.8-27B | 296.0 | 154.2 |
| DeepSeek-Coder-V2-Lite-Instruct | 237.3 | — |
| Qwen3-Coder-Next | 174.3 | — |
| Llama-3.1-8B-Instruct | 158.1 | — |
| Muse-Glimmer-30B | 152.2 | — |
| Qwen3.8-27B-MTP | 112.1 | — |
| Qwen3.6-27B | 101.4 | — |
| Qwen3-14B | 90.8 | — |
| Kimi-Linear-48B-A3B-Instruct | 90.0 | — |
| Qwen3.6-35B-A3B-abliterated-v4 | — | 82.1 |
| Qwen3.6-27B-MTP | 60.1 | — |
| Qwen2.5-Coder-14B-Instruct | 50.3 | — |
| Qwen3-Coder-30B-A3B-Instruct | 26.1 | — |