Can the Apple M4 Max (128GB) run Llama 3.1 70B?
Yes
QuantQ8_0Near-lossless
Memory78.3 GBof 96 GB
Speed~4tokens/s
Yes. The Apple M4 Max (128GB) can run Llama 3.1 70B fully in unified memory at Q8_0 (near-lossless), using about 78.3 GB of the 96 GB available, at roughly 4 tokens/second.
Try other context lengths in the calculator →Memory needed by quantization
At 8,192 tokens of context. macOS lets the GPU use about 96 GB of the 128 GB unified memory by default.
| Quant | Quality | Memory | Fits? | Speed |
|---|---|---|---|---|
| FP16 | Full quality | 144.5 GB | Too big | — |
| Q8_0 | Near-lossless | 78.3 GB | Fits | ~4 tok/s |
| Q6_K | Excellent | 61.2 GB | Fits | ~6 tok/s |
| Q5_K_M | Very good | 53.5 GB | Fits | ~7 tok/s |
| Q4_K_M | Good, most popular | 46.1 GB | Fits | ~8 tok/s |
| Q3_K_M | Noticeably degraded | 37.8 GB | Fits | ~9 tok/s |
| Q2_K | Heavily degraded | 29.8 GB | Fits | ~12 tok/s |
How to run Llama 3.1 70B on the Apple M4 Max (128GB)
With Ollama installed, one command downloads and starts it:
ollama run llama3.1:70bOllama's default download is Q4_K_M (about 46.1 GB here). You have room for Q8_0, which is near-lossless: pick that GGUF in LM Studio or from Hugging Face.
GPUs that run Llama 3.1 70B comfortably
Models that run well on the Apple M4 Max (128GB)
Run it in the cloud
Don't want to buy hardware? Run Llama 3.1 70B on a rented GPU today.
Rent a cloud GPU by the hour. No upfront cost, cancel any time.