Collections

The largest fully-annotated open egocentric human dataset. 100,000 hours across 15,000+ tasks and scenes.

Efficient reasoning and thinking language models for multilingual generation, coding, and AI assistant workflows.

The Qwen3.8-27B-Uncensored family from OrcaRouter 🐋 FP8 ✅ NVFP4 ✅ GGUF ✅ INT8 ✅ MLX ✅ Full precision ✅

Introducing Dynamic V3.0 quants, our new SOTA quantization methodology.