# Inkling: An Open-Weights AI Model Inkling offers a customizable foundation model with multimodal capabilities. - **Source:** [Hacker News](https://news.ycombinator.com) - **Published:** 2026-07-15 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/b9e94cde-ff90-4ef8-adc5-737ea4512dc4 ## Inkling's Mission Inkling embarked on a mission to develop AI that extends human will and judgment. To this end, it created a platform that allows anyone to customize models and introduced an AI system designed for interactive collaboration. ## Open-Weights Model Inkling was released as a model with full weights accessible. This model is a Mixture-of-Experts transformer with 975B total parameters and supports a context window of up to 1M tokens. ## Multimodal Training Inkling was pretrained on 45 trillion tokens of text, images, audio, and video data. This multimodal training enables the model to perform strongly across a wide range of tasks. ## Inkling-Small Preview Inkling-Small was introduced as a lighter-weight model. With 12B active parameters, this model delivers strong performance and low cost using a similar training recipe. ## Generalist Model Inkling was trained across a broad spectrum of tasks, including agentic, coding, instruction-following, visual, and audio processing. This breadth enhances the model's customizability. ## Agentic Coding and Tool Use Inkling performs well among open-weights models in agentic coding and tool use. The model was trained to operate within various coding and agent harnesses. ## Fine-Tuning on Tinker Inkling is available for fine-tuning on the Tinker platform. This allows developers to customize the model according to their specific needs. ## Inkling Playground The Inkling Playground provides an interface for developers to interact with Inkling. This helps in understanding the unique feel of the model. ## Real-World Applications Inkling offers a wide range of capabilities for real-world applications. These capabilities can be combined and improved through fine-tuning. ## Trustworthiness and Safety Inkling demonstrates its performance in important qualities such as trustworthiness and safety, which are crucial for real-world use. ## Customizability Inkling was designed as a customizable foundation model. This allows different users to adapt the model to their specific workflows. ## Future Steps Inkling is the first member of a model family. This family will continue to evolve, offering customizable models for more use cases in the future. ## Sources covering this story - https://thinkingmachines.ai/news/introducing-inkling/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._