tl;dr: Inkling is a new open-weights Mixture-of-Experts model (975B total / 41B active parameters, 1M-token context) trained from scratch on 45T multimodal tokens, with weights available on Hugging Face and fine-tuning support via Tinker. It features controllable thinking effort, native text/image/audio reasoning, and competitive—though not frontier-leading—performance across coding, agentic, and multimodal benchmarks, positioning itself as a broad base for customization rather than a benchmark leader. A smaller 276B/12B-active variant, Inkling-Small, is previewed alongside it.