Apple M3 Ultra (80-core GPU, 512GB) × Qwen2.5 Coder 32B Instruct
Can the Apple M3 Ultra (80-core GPU, 512GB) run Qwen2.5 Coder 32B Instruct?
Yes, usable
Yes. The Apple M3 Ultra (80-core GPU, 512GB) can run Qwen2.5 Coder 32B Instruct at up to F16 (needs ~70.5 GB of 384 GB usable), at an estimated ~17.7 tokens/sec at Q4_K_M.
Speed is estimated from memory bandwidth calibrated with a measured 92.14 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 | 19.4 GB | ✅ | ~22.0 |
| Q4_K_M | 23.4 GB | ✅ | ~17.7 |
| Q5_K_M | 27.0 GB | ✅ | ~15.1 |
| Q6_K | 30.8 GB | ✅ | ~13.0 |
| Q8_0 | 38.8 GB | ✅ | ~10.1 |
| F16 | 70.5 GB | ✅ | ~5.4 |
Fastest hardware for Qwen2.5 Coder 32B Instruct
- NVIDIA GeForce RTX 5090 — ~55.7 tok/s at Q4_K_M
- NVIDIA GeForce RTX 4090 — ~35.8 tok/s at Q4_K_M
- AMD Radeon RX 7900 XTX — ~35.4 tok/s at Q4_K_M
- NVIDIA GeForce RTX 3090 Ti — ~32.9 tok/s at Q4_K_M
- NVIDIA GeForce RTX 3090 — ~30.4 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.