llamaperf

DeepSeek V4.1 Flash 552B (16B active)

on M5 Ultra 512GB · oMLX · 128,000 ctx

Tone: mixedApple hardware
Nov 16, 2026
Throughput
54.0 t/s gen · 1100.0 tokens/s prompt processing (prefill)Prompt processing is how fast the input is read, not how long you wait for the first word. How to measure time to first token.
Quant
oQ4e
System RAM
512 GB

Summary

User reports DeepSeek-V4.1-Flash-1017-oQ4e-mtp at up to 1100 t/s prompt processing and 54 t/s generation on a Mac Studio M5 Ultra with 512GB RAM. Setup is oMLX with an oQ4e quant and MTP speculation, a 128k context window, and engram tables stored on SSD to keep RAM free. Beyond 128k context both prefill and generation drop, and concurrent requests split the maximum generation speed between instances. The user asks whether this is the top speed without changing the model.

Compare this run

Medians of plain runs: one device, one request, no speculative decoding, the whole model in the device's memory. The report above is one setup, so read it beside them and not as one of them.

DeepSeek V4.1 Flash 552B · 16B active at 3-bit by hardware

Bars are medians on one scale, 0 to 73.4 tokens/s. Press ▶ to watch text arrive at a row's speed.

Models most run on the M5 Ultra 512GB

Bars are medians on one scale, 0 to 129.2 tokens/s. Press ▶ to watch text arrive at a row's speed.

Everything reported on the M5 Ultra 512GB →Every DeepSeek V4.1 Flash report →Estimate DeepSeek V4.1 Flash on this card →