Open Models - April 2026 - One of the best months of all time for Local LLMs?

Cuda + ROCm simultaneously with -DGGML_BACKEND_DL=ON !

Follow-up: Qwen3.6-27B on 1× RTX 3090 — pushing to ~218K context + ~50–66 TPS, tool calls now stable (PN12 fix)

New Stealth Model : Owl Alpha

DeepSeek Vision/Multimodal 👀

No, nothing special, just a tiny local language model playing a game it itself wrote.

I stumbled on a Gemma 4 chat template bug for tools and fixed it