Open-Source Model Glimmer Runs at 50 tok/s on a MacBook

Developer TimDarcet revealed details of Glimmer, an open-weight model for local agents, whose pipeline includes pretraining, reasoning midtraining and 131k long context, running at about 50 tok/s on a MacBook with 4-bit quantization.

2026-10-08 ~ 2026-10-08 · 2 related posts

• Glimmer model details: full training pipeline and 50 tok/s on a MacBook via 4-bit quant + DFlash — TimDarcet · 2026-10-08
• Meta's open-weight Glimmer runs local agents at 50 tok/s on a MacBook — TimDarcet · 2026-10-08