M3 Ultra 512GB vs M5 Max 128GB for local LLMs
41 reports on the M3 Ultra 512GB and 48 on the M5 Max 128GB
Which is faster for local LLMs?
On Qwen3.8 125B · 6B active at 4-bit, the M3 Ultra 512GB has 28.0 generation tokens/s (2 runs); prompt processing 558.8 tokens/s (one run); the M5 Max 128GB has 55.0 generation tokens/s (one run). 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 33% higher on the M3 Ultra 512GB (819 GB/s on the M3 Ultra 512GB, 614 on the M5 Max 128GB). Across 3 shared models with generation measurements on both cards, the M3 Ultra 512GB's median is higher on 1 and the M5 Max 128GB's on 2.
- Memory
- 512GB unified
- Memory bandwidth
- 819 GB/s
- FP16 compute
- 28.4 TFLOPS
- Reports
- 41
- Memory
- 128GB unified
- Memory bandwidth
- 614 GB/s
- FP16 compute
- not on record
- Reports
- 48
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 512GB | M5 Max 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active4-bit | 28.02 runsPP 558.8 (one run)llama.cpp, oMLX250K context, 1 not stated | 55.0one runengine not statedcontext not stated |
| DeepSeek V4 Flash 284B · 13B active4-bit | 29.9one runPP 426.5 (one run)MLXcontext not stated | 36.1one runPP 498.3 (one run)oMLX1K context |
| Qwen3.6 35B · 3B active8-bit | 81.2one runPP 2324.6 (one run)engine not statedcontext not stated | 79.4one runPP 2153.0 (one run)llama.cpp125K context |
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 512GB | M5 Max 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active | 13.6 t/sFP16 | 41.2 t/sQ5_K_S |
| DeepSeek V4 Flash 284B · 13B active | 22.6 t/sQ8_0 | 64.8 t/sQ2_K |
| Qwen3.6 35B · 3B active | 28.7 t/sFP16 | 28.5 t/sFP16 |
| Qwen3.8 27B | 8.0 t/sFP16 | 9.0 t/sFP16 |
| Qwen3.6 27B | 8.0 t/sFP16 | 9.0 t/sFP16 |
| Gemma 4 26B · 4B active | 21.5 t/sFP16 | 24.4 t/sFP16 |
| DeepSeek V4.1 Flash 552B · 16B active | 30.0 t/sQ4_K_M | doesn't fit at these settings |
| GLM-5.3 320B · 18B active | 7.4 t/sQ8_0 | doesn't fit at these settings |
| Qwen3.5 9B | 23.9 t/sFP16 | 27.1 t/sFP16 |
Change the context, RAM or model in the calculator: M3 Ultra 512GB · M5 Max 128GB
Every report on each card, with its source: M3 Ultra 512GB · M5 Max 128GB