Gemma 4 running inside Godot via Vulkan compute shaders

A Reddit user successfully ran Gemma 4 natively inside Godot 4.7 using only GDScript and Vulkan compute shaders, without llama.cpp, Python, or external servers. The project loads a Q4_K_M quantized GGUF file and handles tokenization and sampling entirely in-engine.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Plaud unveils 4G-connected AI earbuds for on-the-go tasks
- Adobe adds AI Assisted Editor interface to Photoshop
- Webinar: Build AI threat readiness for security ops
- Anthropic's Mike Krieger: Claude ported 100k lines of Python to TypeScript in a weekend
- OpenAI to show ads on ChatGPT free and Go tiers in India