Back to home

Apple Silicon

8 articles tagged with this topic

QwenApple Silicon

Qwen 3.8 Tested: Deep Thinking Burns 5.5x Tokens—Local Deployment Math Changes

Reddit user tested Qwen3.8-27B on M5 Max: deep thinking uses 5.5x tokens, 6x time; disabling tanks quality. The "thinking" cost gap is exposed.

4h ago2 min read
AppleMac mini

Mac mini's M6 upgrade drops the bar for running AI models locally

Apple's Mac mini refresh with next-gen Apple Silicon signals local AI is moving from geek toy toward semi-mainstream — but production readiness still

3d ago2 min read
h3.cApple Silicon

33B Audio-Video Model on Apple Silicon: Open Source Tallies Acceleration's True Cost

This week, h3.c ported a 33B audio-video model to Apple Silicon, exposing six distinct optimization knobs instead of a single "fast=true" switch.

Aug 222 min read
QwenApple Silicon

Qwen Runs 45 Tokens/Second Locally — Apple Silicon's Silent Win for Open LLMs

Reddit user benchmarked Qwen 3.8B at 45+ tok/s on Apple silicon — a record on consumer hardware. Local LLM inference shifts from geek toy to viable op

Aug 212 min read
QwenApple Silicon

Apple Silicon Runs Qwen 27B 3x Faster — Local LLMs Enter Usable Territory

mlx-dspark ports DeepSeek's speculative decoding to Apple Silicon, giving Qwen 27B a 3x speedup on M-series Macs with no quality loss. Local LLMs near

Aug 152 min read
MetaMuse Glimmer

Meta's Muse Glimmer 30B Hits 3.3x Speedup on Mac as Open Source Closes the Gap

Developer A-Rahim uses speculative decoding to push Meta's Muse Glimmer 30B up to 3.3x faster on M4 Pro, with output exactly matching the original.

Aug 122 min read
Anubis-OSSApple Silicon

Local AI Gets Serious: Anubis-OSS Leaderboard Tracks 218 Models, 10 Apple Chips

Anubis-OSS leaderboard updates: 371 submissions, 218 models, 10 Apple chips. This data proves local open-source model deployment is no longer a geek t

May 52 min read
llama.cppQwen

37 LLMs Benchmarked on MacBook Air M5 32GB: Full Speed Results

Community benchmark of 37 local LLMs on M5 Air 32GB using llama-bench reveals MoE models as clear winners for speed-to-quality ratio.

Apr 62 min read