Moe

The 2.8T MoE model Kimi K3 requires 1.45 TB of memory, exceeding any laptop's RAM. The Deltafin fork solves this by streaming expert weights from NVMe SSDs to RAM layer by layer. This article breaks down benchmarks on a 128GB M5 Max MacBook Pro, revealing true decode rates, SSD scaling, and practical limitations for local inference.

From 8 GB to 512 GB VRAM: which local AI models run on your hardware — with the production recipes and benchmarks from Mia's AI Lab.

DeepSeek-V4-Pro-0813 went GA on August 13, posting a Terminal-Bench 2.1 score of 87.9% — the first open-weight model to beat a frontier closed model on that benchmark. But the GA weights aren't on Hugging Face, a 264% price hike hits August 16, and several self-reported benchmarks still await independent replication.