Modular's Newly Open-Sourced MAX Beats llama.cpp on Mobile CPU and GPU, Benchmarks Reproducible
clattner_llvm · x · 2026-10-28
Chris Lattner highlights that Modular's freshly open-sourced MAX is already showing strong results: engineer Apurva published a detailed post covering bring-up, optimizations, failed experiments, and benchmark results with reproducible source code.
MAX reportedly beats llama.cpp on both mobile CPU and GPU, supports multiple GPU vendors, and handles the device/desktop split properly. Lattner calls it just the start and invites scrutiny of the measurements.
More from Infra
- NVIDIA open-sources SimReady: 2,000+ simulation-ready 3D assets top Hugging Face charts — mishig25 · 2026-10-28
- Together GPU Clusters adds auto repair: from GPU fault to healthy node with no human in the loop — togethercompute · 2026-10-28
- Supabase launches Compute in private alpha: run AI agents and long-running jobs next to your Postgres data — dshukertjr · 2026-10-28
- Atreides' Gavin Baker: operating cash flow will fund more of the 2027 AI buildout than expected — BenBajarin · 2026-10-28
- Anthropic built Scion, its own source control, as daily merged changes grew from 2,500 to 30,000 — krishnan · 2026-10-28
- fal.ai publicly hunting for B300/GB300 compute, 1MW+ chunks, delivery before year-end — isidentical · 2026-10-28