JustVugg/colibri

Runs frontier MoE models up to 2.8T parameters on consumer hardware using pure C with zero dependencies, treating storage, RAM, and VRAM as a single unified memory tier.

View on index · View in 3D Map

Related repos

• Run large models on Mac without RAM limits
• open source video generation for low VRAM GPUs
• Runs 405B Llama3.1 on 8GB VRAM
• Runs hundreds of large models on a single GPU by loading them from SSD to VRAM up to 10x faster than alternative loaders, achieving coldstarts under 2 seconds.
• Runs large language models on iOS and macOS
• Cross-platform database client built with Rust and Svelte
• Framework for RAG pipelines using MCP architecture
• Static single-batch CUDA inference engine
• CPU simulation for flaky-free benchmarks
• Persistent memory system for Hermes Agent