🤖 New Multimodal Model Inkling Introduced
Thinking Machines Lab has released Inkling — a multimodal open-weights model based on the Mixture-of-Experts (MoE) architecture. The main model has 975 billion parameters (41 billion active), while the Inkling-Small version has 276 billion parameters (12 billion active). The model supports a context window of up to 1 million tokens and works natively with audio and video.
🌍 The release of a large open-weights MoE model intensifies competition with proprietary solutions. An innovative approach to managing "thinking effort" allows for balancing accuracy and generation costs.
👤 Users receive an open model capable of adjusting the depth of reasoning to specific tasks and budgets, while also demonstrating capabilities in self-learning and agentic coding.
Source 1: https://thinkingmachines.ai/news/introducing-inkling/