M3 Ultra 256GB vs AMD Strix Halo 128GB for local LLMs
15 reports on the M3 Ultra 256GB and 320 on the AMD Strix Halo 128GB
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 AMD Strix Halo 128GB has 24.5 generation tokens/s (6 runs); prompt processing 1700.5 tokens/s (5 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 3.2x higher on the M3 Ultra 256GB (819 GB/s on the M3 Ultra 256GB, 256 on the AMD Strix Halo 128GB). Across 3 shared models with generation measurements on both cards, the M3 Ultra 256GB's median is higher on 3 and the AMD Strix Halo 128GB's on 0.
- Memory
- 256GB unified
- Memory bandwidth
- 819 GB/s
- FP16 compute
- 28.4 TFLOPS
- Reports
- 15
- Memory
- 128GB unified
- Memory bandwidth
- 256 GB/s
- FP16 compute
- 59 TFLOPS
- Reports
- 320
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 | AMD Strix Halo 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active4-bit | 25.7one runengine not statedcontext not stated | 24.56 runsPP 1700.5 (5 runs)Gufo, llama.cpp, 1 not stated128K to 160K context, 3 not stated |
| Qwen3.8 27B4-bit | 14.0one runPP 93.1 (one run)Ollamacontext not stated | 13.14 runsPP 378.2 (3 runs)llama.cpp, 1 not stated32K context, 3 not stated |
| Qwen3.5 9B4-bit | 107.5one runengine not statedcontext not stated | 31.1one runPP 819.2 (one run)engine 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 | AMD Strix Halo 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active | 27.3 t/sQ8_0 | 12.6 t/sQ6_K |
| Qwen3.8 27B | 8.0 t/sFP16 | 3.5 t/sFP16 |
| Qwen3.5 9B | 23.9 t/sFP16 | 10.5 t/sFP16 |
| Qwen3.6 35B · 3B active | 28.7 t/sFP16 | 12.6 t/sFP16 |
| DeepSeek V4 Flash 284B · 13B active | 36.0 t/sQ4_K_M | 33.5 t/sQ2_K |
| Qwen3.6 27B | 8.0 t/sFP16 | 3.5 t/sFP16 |
| Gemma 4 26B · 4B active | 21.5 t/sFP16 | 9.5 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 | 11.1 t/sQ2_K |
Change the context, RAM or model in the calculator: M3 Ultra 256GB · AMD Strix Halo 128GB
Every report on each card, with its source: M3 Ultra 256GB · AMD Strix Halo 128GB