M3 Ultra 256GB vs NVIDIA RTX Pro 6000 Blackwell for local LLMs
15 reports on the M3 Ultra 256GB and 70 on the NVIDIA RTX Pro 6000 Blackwell
Which is faster for local LLMs?
On Qwen3.8 125B · 6B active at 4-bit, the M3 Ultra 256GB has 25.7 generation tokens/s (one run); the NVIDIA RTX Pro 6000 Blackwell has 105.5 generation tokens/s (6 runs); prompt processing 5110.0 tokens/s (6 runs). Those are community reports whose engines and settings differ, so the gap is a rough sign, not a controlled measure of the two cards. Memory bandwidth, which caps how fast a card can write, is 2.2x higher on the NVIDIA RTX Pro 6000 Blackwell (819 GB/s on the M3 Ultra 256GB, 1,792 on the NVIDIA RTX Pro 6000 Blackwell). Across 2 shared models with generation measurements on both cards, the M3 Ultra 256GB's median is higher on 0 and the NVIDIA RTX Pro 6000 Blackwell's on 2.
- Memory
- 256GB unified
- Memory bandwidth
- 819 GB/s
- FP16 compute
- 28.4 TFLOPS
- Reports
- 15
- Memory
- 96GB
- Memory bandwidth
- 1,792 GB/s
- FP16 compute
- 503 TFLOPS
- Reports
- 70
Measured on both
Generation and prompt processing speeds from plain runs (one device, one request, no speculative decoding, the whole model in the device's memory) of the same model size at the same bit class. Each side shows the engines and reported context behind its median. Context can mean a configured window or prompt depth, so these are community comparisons with differing settings. How to read these
| Model | M3 Ultra 256GB | NVIDIA RTX Pro 6000 Blackwell |
|---|---|---|
| Qwen3.8 125B · 6B active4-bit | 25.7one runengine not statedcontext not stated | 105.56 runsPP 5110.0 (6 runs)llama.cpp, ik_llama.cpp +2, 3 not stated512 to 256K context, 1 not stated |
| Qwen3.8 27B4-bit | 14.0one runPP 93.1 (one run)Ollamacontext not stated | 67.0one runengine not statedcontext not stated |
Estimated by the calculator
The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape. Where a model was also measured above, the measurement wins.
| Model | M3 Ultra 256GB | NVIDIA RTX Pro 6000 Blackwell |
|---|---|---|
| Qwen3.8 125B · 6B active | 27.3 t/sQ8_0 | 83.1 t/sQ5_K_S |
| Qwen3.8 27B | 8.0 t/sFP16 | 18.3 t/sFP16 |
| Qwen3.6 35B · 3B active | 28.7 t/sFP16 | 57.5 t/sFP16 |
| DeepSeek V4 Flash 284B · 13B active | 36.0 t/sQ4_K_M | 130.8 t/sQ2_K |
| Qwen3.6 27B | 8.0 t/sFP16 | 18.3 t/sFP16 |
| Gemma 4 26B · 4B active | 21.5 t/sFP16 | 49.3 t/sFP16 |
| DeepSeek V4.1 Flash 552B · 16B active | 48.0 t/sQ2_K | doesn't fit at these settings |
| GLM-5.3 320B · 18B active | 19.1 t/sQ3_K_M | 6.8 t/sQ2_K, experts in RAM |
Change the context, RAM or model in the calculator: M3 Ultra 256GB · NVIDIA RTX Pro 6000 Blackwell
Every report on each card, with its source: M3 Ultra 256GB · NVIDIA RTX Pro 6000 Blackwell