Local AI Guide

See the speed before choosing hardware

Prefill and token generation speed

Qwen3.8 27B on Mac Studio M4 Max 128GB

Check prefill and generation for Qwen3.8 27B on a Mac Studio M4 Max 128GB. Compare whether you need the large memory capacity or a smaller configuration is enough for this model.

Benchmark-based estimate · 4K input · one user

Last updated

27.9 ~ 60.4 tok/sDecode speed(baseline: No matching baseline)
Model fully loaded (17.86GB / 114GB)
First token (TTFT)
5.8 to 10.5 secondsContext 4,096 input tokens included
Prefill speed
393 ~ 707 tok/s5.8 to 10.4 seconds (Benchmark-calibrated)
Total time (E2E)
7.7 to 14.6 seconds117 output tokens included
Press Play to process the input and generate the answer.
Ready
Change device or model
546 GB/s
27B
Expand settings 4-bit · 4K · MTP acceleration (experimental)
QuantizationNeed 17.86GB
Input context4K tokens
MTP acceleration (experimental)
Baseline
Sample question:
117 output tokens

Reading this result

Unused memory does not make a response proportionally faster. Separate plans for larger models or concurrent work from running a single 27B model today.

Required / available memory
17.9GB / 114GB
Runtime
oMLX 0.6.1 · native MTP k=3 · ANE prefill not enabled in this recipe

Compare this model side by side

You can inspect the same model on an M4 Pro 48GB. Decide whether extra memory headroom, rather than the speed difference, is what your work needs.

Include prefillCompare generation only

Check configuration

Studio M4 Max 128GB · Memory 128GB

Qwen3.8 27B Q4_K_M configuration. Check that the retailer lists the same memory option.

What changes with a longer document?Expand input length, waiting time, speed and memory
Detailed analysis and all 61 hardware configurations

Popular configurations

Speed simulations by device and model

View all 20

Qwen3.8 27B on Mac mini M4 32GB: response speed

5.1 ~ 5.7 tok/s · 16.4 ~ 35.5 s

Qwen3.8 27B: is a Mac mini M4 Pro 48GB worth it?

12.3 ~ 13.7 tok/s · 7.8 ~ 14.3 s

Qwen3.8 27B on MacBook Pro M5 Pro 64GB

14.7 ~ 17.1 tok/s · 8.6 to 10.0 seconds

Qwen3.8 27B on RTX 3090: do you need an upgrade?

42.3 ~ 46.9 tok/s · 4.4 to 7.5 seconds