4x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open at 36.7 tok/s
| GPU(s) | 4x Apple M5 Ultra (unified, Mac Studio) |
|---|---|
| VRAM / unified memory (GB) | |
| CPU | 4x Apple M5 Ultra |
| Motherboard | |
| RAM | |
| PSU | |
| Other | Four Mac Studios in a Thunderbolt 5 mesh, two cables per pair (RDMA over Thunderbolt, MLX jaccl backend) |
| Reported total cost | |
| Power |
Measured speed (generation tokens/sec)
📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
| Model | Quant | Engine | tok/s | Details |
|---|---|---|---|---|
| Muse Spark Open 816B-A42B | NVFP4, 8,192-token context | MLX, 4-node tensor parallel | 36.73 | |
| Mistral Medium 3.5 128B | FP8, 8,192-token context | MLX, 4-node tensor parallel | 23.37 | |
| Qwen3.8 27B | BF16, 8,192-token context | MLX, 4-node tensor parallel | 50.89 |
Apple Machine Learning Research, 15 Oct 2026 (vendor benchmark, single user, 128 generated tokens). Apple does not state the memory of each Mac in the cluster runs.
Source: https://machinelearning.apple.com/research/mlx-distributed-inference
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)
Quick answers: Can 4x Apple M5 Ultra run Mistral Medium 3.5 128B? · Can 4x Apple M5 Ultra run Muse Spark Open 816B-A42B? · Can 4x Apple M5 Ultra run Qwen3.8 27B?