Localmaxxing | Tomasz Tunguz

kept by eddie may 12

About half of agent tasks succeed on a local 35B model, with 2.1x faster latency than cloud models, making local inference advantageous for routine work.

M5 Pro and M5 Max are surprisingly big departures from older Apple Silicon - Ars Technica
Apple's M5 Pro and M5 Max chips introduce a chiplet-based Fusion Architecture and a third CPU core type, splitting CPU/GPU across two dies for the first time in non-Ultra chips.
also on #macbook-pro
How to run Qwen 3.5 locally
Hacker News commenters share experiences and benchmarks for running Qwen 3.5 locally, covering hardware choices, quantization trade-offs, and practical coding performance.
also on Qwen, #benchmarks, #ai, #developer-tools
8 plots that explain the state of open models
Qwen dominates open-model downloads and finetunes globally, while DeepSeek leads in large-scale models and GPT-OSS is the only Western contender gaining adoption.
also on Qwen, #benchmarks, #ai, #developer-tools