llama.cpp v0.4.0: lazy tensor reading, per-slot context limits, video input
llama.cpp shipped v0.4.0 on Sep 4, adding on-demand lazy tensor reading, per-slot server context limits, video input options, and a ggml bump from 0.22.0 to 0.23.0, alongside new model support for Qwen3.8-Flash-Next and Nemotron-3-Puzzle. The same day saw ten tagged builds including Metal fa-vec tunings for M3 Max and OpenCL Adreno SDPA paths.
Why it matters
Lazy tensor loading cuts cold-start memory for large models on consumer hardware, a meaningful win for local-inference users.
Importance: 3/5
Major version bump of the reference local-inference engine