Inkling: Our Open-Weights Model
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
A 975B parameter Mixture-of-Experts transformer model, Inkling, is available with full weights for customization, supporting a context window of up to 1M tokens and multimodal reasoning over text, images, and audio. This enables shipping multimodal and highly customizable AI models with potentially lower costs and latency using the lighter Inkling-Small variant, and allows for accessible fine-tuning on Tinker.
975B-parameter open-weights model with 41B active parameters and 1M-token context, pretrained on 45T multimodal tokens. This lets you fine-tune a single model that handles text, images, and audio at scale without hitting proprietary walls, cutting per-token cost and latency while keeping customization simple on Tinker. If you’re shipping agents that need to reason across modalities or adapt quickly to new tasks, Inkling gives you a production-ready base that’s cheaper to run than closed alternatives and easier to tweak than training from scratch.