llamaperf

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

on 2× M5 Ultra 256GB · oMLX · 8,145 ctx

Apple hardware
Nov 19, 2026
Throughput
44.4 t/s gen
System RAM
256 GB

Summary

User benchmarks DeepSeek V4.1 Flash on two M5 Ultra 256GB Macs with replicated attention, reporting 44.39 decode tok/s per user and 3,136 effective prefill tok/s per user at one user. Setup is oMLX 0.8.0 with native weights, identical 8,145-token prompts, 512-token replies and zero cache reuse; prefill, decode and TTFT are per-user medians over three timed rounds after warmup. Total output throughput is given as a three-run range of 33.70–36.51 tok/s, and TTFT is 2.60 s.

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 256GB

Bars are medians on one scale, 0 to 141.0 tokens/s. Press ▶ to watch text arrive at a row's speed.

Everything reported on the M5 Ultra 256GB →Every DeepSeek V4.1 Flash report →Estimate DeepSeek V4.1 Flash on this card →