Observed arrival · 2026-10-27
AX Engine’s Mac inference benchmark board
A Mac-first Apple Silicon inference runtime that uses speculative decoding and offers a self-hosted, OpenAI-compatible endpoint.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Apple Silicon users running local language models
- Worth noticing
- The benchmark labels itself campaign evidence, not certification, and reports medians across 20 runs on one M5 Max.
Field notes
The comparison panels separate dense, mixture-of-experts, and Gemma workloads rather than presenting one blended speed ranking. They also disclose when a model pack was unavailable to a comparator, and distinguish a separate Tiel completion test from default-server speed.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue