M2 Ultra 128GB vs AMD Strix Halo 128GB for local LLMs
4 reports on the M2 Ultra 128GB and 320 on the AMD Strix Halo 128GB
Which is faster for local LLMs?
On Qwen3-Coder 30B · 3B active at 4-bit, the M2 Ultra 128GB has 70.3 generation tokens/s (2 runs); prompt processing 1088.7 tokens/s (2 runs); the AMD Strix Halo 128GB has 69.9 generation tokens/s (6 runs); prompt processing 401.5 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 M2 Ultra 128GB (819 GB/s on the M2 Ultra 128GB, 256 on the AMD Strix Halo 128GB).
- Memory
- 128GB unified
- Memory bandwidth
- 819 GB/s
- FP16 compute
- 27.2 TFLOPS
- Reports
- 4
- 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 | M2 Ultra 128GB | AMD Strix Halo 128GB |
|---|---|---|
| Qwen3-Coder 30B · 3B active4-bit | 70.32 runsPP 1088.7 (2 runs)llama.cppcontext not stated | 69.96 runsPP 401.5 (5 runs)llama.cpp8K context, 5 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 | M2 Ultra 128GB | AMD Strix Halo 128GB |
|---|---|---|
| Qwen3-Coder 30B · 3B active | 29.2 t/sFP16 | 12.6 t/sFP16 |
| Qwen3.8 27B | 8.3 t/sFP16 | 3.5 t/sFP16 |
| Qwen3.8 125B · 6B active | 37.9 t/sQ5_K_S | 12.6 t/sQ6_K |
| Qwen3.6 35B · 3B active | 26.2 t/sFP16 | 12.6 t/sFP16 |
| DeepSeek V4 Flash 284B · 13B active | 59.6 t/sQ2_K | 33.5 t/sQ2_K |
| Qwen3.6 27B | 8.3 t/sFP16 | 3.5 t/sFP16 |
| Gemma 4 26B · 4B active | 22.5 t/sFP16 | 9.5 t/sFP16 |
Change the context, RAM or model in the calculator: M2 Ultra 128GB · AMD Strix Halo 128GB
Every report on each card, with its source: M2 Ultra 128GB · AMD Strix Halo 128GB