llamaperf

Inkling-2 Mini

on M4 Pro 48GB · oMLX · 30,000 ctx

Apple hardware
Nov 22, 2026
Throughput
23.9 t/s gen · 185.3 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
48 GB

Summary

User reports Inkling-2 Mini at 185.3 t/s prompt processing and 23.9 t/s generation on an M4 Pro 48GB at 30k context. Setup is oMLX 0.8.2 with an oQ4e quant and MTP speculation, thinking mode off. For comparison, the user reports Qwen3.8 27B on the Splash engine at 216.1 t/s prompt and 40 t/s generation.

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.

Models most run on the M4 Pro 48GB

Bars are medians on one scale, 0 to 8.2 tokens/s. Press ▶ to watch text arrive at a row's speed.

Everything reported on the M4 Pro 48GB →Every Inkling-2 report →Estimate Inkling-2 on this card →