GPU comparison
Mac Studio M3 Ultra (96GB) vs NVIDIA RTX Pro 6000 Blackwell (96GB) for local LLMs
How the two stack up for running open-source LLMs locally — memory, bandwidth, price, and how many of the 64 tracked models each one can run.
| Mac Studio M3 Ultra (96GB) | NVIDIA RTX Pro 6000 Blackwell (96GB) | |
|---|---|---|
| Memory | 96.0 GB | 96.0 GB |
| Bandwidth | 819 GB/s | 1792 GB/s |
| Price (approx) | $5,299 | $12,999 |
| LLMs it runs | 55 of 64 | 57 of 64 |
| Best model it runs | Qwen3.8 27B · 17–25 tok/s | Qwen3.8 Flash Next (MoE) · 205–307 tok/s |
The NVIDIA RTX Pro 6000 Blackwell (96GB) runs 2 more of the tracked models (57 vs 55). With the same nominal memory, more of it is usable for model weights on this architecture. The NVIDIA RTX Pro 6000 Blackwell (96GB) has more memory bandwidth (1792 GB/s), so it generates tokens faster at the same model and quant.