← Dashboard  /  News

2026-10-08

AI/Tech Digest 2026-10-08

AI & local LLM

1. llama : add a GPU cache for MoE experts kept in host memory by am17an · Pull Request #29887 · ggml-org/llama.cpp [LAB]
   This PR introduces a GPU cache for MoE experts in host memory, a critical optimization for running large MoE models on hardware with limited VRAM.
   https://www.reddit.com/r/LocalLLaMA/comments/1x03xkc/llama_add_a_gpu_cache_for_moe_experts_kept_in/ (reddit.com)
2. New LFM to be released today [LAB]
   Liquid AI is releasing new Liquid Foundation Models (LFM), which are significant for tracking new architectural approaches to model scaling.
   https://www.reddit.com/r/LocalLLaMA/comments/1wzuto0/new_lfm_to_be_released_today/ (reddit.com)
3. Claude Haiku 5.5 [LAB]
   A new frontier model release from Anthropic, relevant for tracking the performance and efficiency of high-speed, low-cost model generations.
   https://simonwillison.net/2026/Oct/7/claude-haiku-5-5/ (simonwillison.net)