Hardware › Mac Ultra local LLM benchmarks: 10 measured configs
Mac Ultra local LLM benchmarks: 10 measured configs
10 measured local-LLM configs on Apple Mac Studio (M-series Ultra); fastest 14B+ model Qwen3.6-35B-A3B at 109.1 tok/s. Official specs, Singapore prices read daily, every number linked to its source.
Snapshot 2026-10-16 · 394 configs on AI boxes and 423 on GPU cards · 123 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.
Silicon: Apple M-series Ultra — Mac Studio. Official specs.
Quick answers
- What is the fastest model on Mac Ultra?
- Qwen3.6-35B-A3B at 109.1 tok/s single-stream (MLX-4BIT, mlx), source last checked 2026-09-29. Models under 14B parameters are left out of this answer. source
- What is the fastest 100B+ model on Mac Ultra?
- Qwen3.8-Flash-Next at 76.4 tok/s single-stream (MLX-4BIT, mlx (speculative decode, mtp)), source last checked 2026-09-06. source
- What is the largest model measured on Mac Ultra?
- Qwen3.8-Flash-Next (177B parameters) at 76.4 tok/s single-stream (MLX-4BIT, mlx (speculative decode, mtp)), source last checked 2026-09-06. source
- How many configs does TokenMark track for Mac Ultra?
- 10 single-stream configs across 9 models, each linked to the README or model card it came from.
Official specs
- current lineup: M5 Ultra (2026 Mac Studio, announced 2026-08-25, available 2026-09-22); previous generation M3 Ultra (2025) is what most published benchmarks were measured on (source)
- form factor: desktop, 3.7 in (9.5 cm) H x 7.7 in (19.7 cm) W x 7.7 in (19.7 cm) D (source)
- max continuous power w: 480 (source)
- wireless: Apple N1 chip: Wi-Fi 7 (802.11be), Bluetooth 6, Thread (source)
- memory bandwidth gbps: 1200 (source)
- max unified memory gb: 512 (source)
Where to buy Apple Mac Studio (M-series Ultra) in Singapore
Read 17 Oct 2026 (SGT) from each seller's own product page. Prices in SGD. Overseas stores are converted at USD/SGD 1.2792 with 9% GST at import added; shipping is not included. Singapore sellers already include GST. List prices on the day, not a quote. No affiliate links.
Apple Mac Studio, M5 Ultra (96 GB base)
- Apple (SG) (Official store, configure to order): S$7,999 to S$9,949
Measured single-stream configs on Apple Mac Studio (M-series Ultra)
| Model | Quant | Backend | Decode | Context | Trust | Source |
|---|---|---|---|---|---|---|
| Qwen3.6-35B-A3B | MLX-4BIT | mlx | 109.1 tok/s (prefill 6012) | 32000 | Extracted, source-linked | source |
| Qwen3.8-Flash-Next | MLX-4BIT | mlx (speculative decode, mtp) | 76.4 tok/s | Extracted, source-linked | source | |
| Qwen3.5-122B-A10B | MLX-4BIT | mlx | 65.3 tok/s (prefill 2300) | 32000 | Extracted, source-linked | source |
| Qwen3-14B | MLX-4BIT | mlx | 55.4 tok/s (prefill 2278) | 32000 | Extracted, source-linked | source |
| GLM-5.3-Flash | mlx (speculative decode, mtp) | 55.2 tok/s | Extracted, source-linked | source | ||
| Qwen3.8-27B | MLX-4BIT | mlx | 42.6 tok/s (prefill 1526) | 32000 | Extracted, source-linked | source |
| GLM-5.3-Flash | MLX-4BIT | mlx | 34.2 tok/s | Extracted, source-linked | source | |
| DeepSeek-V4-Flash-0731-MXFP4-MLX-Abliterated | MXFP4 | mlx | 29.6 tok/s (prefill 454) | 1048576 | Extracted, source-linked | source |
| GLM-5.3-Flash Abliterated | MLX-4BIT | mlx (speculative decode, mtp) | 24 tok/s (prefill 365) | 16384 | Extracted, source-linked | source |
| Gemma-4-31B | BF16 | mlx | 13.6 tok/s (prefill 1217) | 32000 | Extracted, source-linked | source |
More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.