M3 Ultra 512GB vs M4 Max 128GB for local LLMs
41 reports on the M3 Ultra 512GB and 53 on the M4 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 M4 Max 128GB has 57.2 generation tokens/s (2 runs); prompt processing 586.0 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 1.5x higher on the M3 Ultra 512GB (819 GB/s on the M3 Ultra 512GB, 546 on the M4 Max 128GB). Across 5 shared models with generation measurements on both cards, the M3 Ultra 512GB's median is higher on 2 and the M4 Max 128GB's on 3.
- Memory
- 512GB unified
- Memory bandwidth
- 819 GB/s
- FP16 compute
- 28.4 TFLOPS
- Reports
- 41
- Memory
- 128GB unified
- Memory bandwidth
- 546 GB/s
- FP16 compute
- 18.4 TFLOPS
- Reports
- 53
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 | M4 Max 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active4-bit | 28.02 runsPP 558.8 (one run)llama.cpp, oMLX250K context, 1 not stated | 57.22 runsPP 586.0 (one run)oMLX, 1 not statedcontext not stated |
| Gemma 4 26B · 4B active4-bit | 87.8one runPP 2595.3 (one run)llama.cppcontext not stated | 88.1one runPP 1408.1 (one run)llama.cppcontext not stated |
| Qwen3.5 122B · 10B active4-bit | 49.5one runPP 784.5 (one run)MLXcontext not stated | 15.6one runllama.cppcontext not stated |
| Qwen3.8 125B · 6B active8-bit | 25.7one runllama.cpp32K context | 60.0one runPP 730.0 (one run)mlx-servecontext not stated |
| Qwen3.8 27B4-bit | 31.8one runPP 391.2 (one run)MLXcontext not stated | 21.8one runPP 236.6 (one run)llama.cppcontext 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 512GB | M4 Max 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active | 13.6 t/sFP16 | 30.7 t/sQ5_K_S |
| Gemma 4 26B · 4B active | 21.5 t/sFP16 | 18.2 t/sFP16 |
| Qwen3.5 122B · 10B active | 8.3 t/sFP16 | 18.9 t/sQ5_K_S |
| Qwen3.8 27B | 8.0 t/sFP16 | 6.7 t/sFP16 |
| Qwen3.6 35B · 3B active | 28.7 t/sFP16 | 21.2 t/sFP16 |
| DeepSeek V4 Flash 284B · 13B active | 22.6 t/sQ8_0 | 48.3 t/sQ2_K |
| Qwen3.6 27B | 8.0 t/sFP16 | 6.7 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 | 20.2 t/sFP16 |
Change the context, RAM or model in the calculator: M3 Ultra 512GB · M4 Max 128GB
Every report on each card, with its source: M3 Ultra 512GB · M4 Max 128GB