Apple M1 Max (32-core GPU, 64GB) × Llama 3.2 3B Instruct
Can the Apple M1 Max (32-core GPU, 64GB) run Llama 3.2 3B Instruct?
Yes
Yes. The Apple M1 Max (32-core GPU, 64GB) can run Llama 3.2 3B Instruct at up to F16 (needs ~8.3 GB of 48.0 GB usable), at an estimated ~121 tokens/sec at Q4_K_M.
Speed is estimated from memory bandwidth calibrated with a measured 61.19 tok/s Llama 2 7B Q4_0 result (source). No direct public benchmark of this exact pair is recorded yet.
By quantization (8K context)
| Quant | Memory needed | Fits? | Est. tok/s |
|---|---|---|---|
| Q3_K_M | 3.3 GB | ✅ | ~150 |
| Q4_K_M | 3.7 GB | ✅ | ~121 |
| Q5_K_M | 4.1 GB | ✅ | ~103 |
| Q6_K | 4.5 GB | ✅ | ~88.6 |
| Q8_0 | 5.2 GB | ✅ | ~68.8 |
| F16 | 8.3 GB | ✅ | ~36.5 |
Fastest hardware for Llama 3.2 3B Instruct
- NVIDIA GeForce RTX 5090 — ~300 tok/s at Q4_K_M
- NVIDIA GeForce RTX 4090 — ~300 tok/s at Q4_K_M
- NVIDIA GeForce RTX 3090 — ~300 tok/s at Q4_K_M
- NVIDIA GeForce RTX 3090 Ti — ~300 tok/s at Q4_K_M
- NVIDIA GeForce RTX 5080 — ~300 tok/s at Q4_K_M
No suitable GPU? Rent one by the hour on Vast.ai or RunPod (affiliate links).
Get new models & benchmarks in your inbox
No spam. Unsubscribe anytime by replying to any email.