Gemma 4 launch hits LM Studio bugs

The rapid release of optimized quants by Unsloth has outpaced the stability of local inference engines for Gemma 4's frontier architecture. Gemma 4 E4B is the first small-scale model with native audio and multimodal support, complicating initial GGUF implementations in local tools. While most "failed to load" errors are resolved by upgrading to LM Studio v0.4.8+ and manually refreshing runtimes, the massive 256K context window on larger variants (26B/31B) is causing VRAM allocation crashes on consumer hardware. Unsloth’s day-zero support for Q5_K_M quants confirms their dominance in the fine-tuning pipeline, but local developers should initially limit context length to 8192 to verify successful loading before attempting to utilize the full 256K token capacity.