Thinking Machines Releases Open Multimodal Model Inkling

Thinking Machines Lab has open-sourced Inkling, a massive 975-billion-parameter multimodal AI model that promises to challenge industry leaders like Anthropic and OpenAI while prioritizing low cost and resistance to censorship. The model, trained on vast amounts of video and audio data in addition to text, represents a significant shift toward more accessible and customizable enterprise AI. What sets Inkling apart is its focus on enterprise autonomy. Unlike cloud-dependent models from major providers, Inkling is designed to run on-premises, giving organizations complete control over their data and AI operations. This addresses growing concerns about data privacy and vendor lock-in that have driven many companies to seek alternatives to proprietary AI services. The model's multimodal capabilities are particularly noteworthy. By training on video and audio, Inkling can understand and generate content across multiple formats, making it suitable for tasks ranging from video analysis to voice-based customer service. Its 975-billion-parameter size places it among the largest open models ever released, rivaling the capacity of closed-source competitors. Perhaps most controversially, Thinking Machines Lab has explicitly designed Inkling with "resistance to censorship" as a core feature. The company argues that enterprises should have the freedom to customize their AI's behavior without the content restrictions imposed by major providers. This has drawn both praise from advocates of open AI and criticism from safety researchers concerned about potential misuse. For enterprises, Inkling offers a compelling value proposition: lower operational costs compared to API-based services, full data sovereignty, and the ability to fine-tune the model for specific industry needs. Early adopters are exploring use cases in healthcare, finance, and manufacturing, where sensitive data cannot be sent to external servers. The open-source release includes model weights, training code, and documentation, allowing organizations to deploy Inkling on their own infrastructure. Thinking Machines Lab is also offering commercial support and customization services, positioning itself as a viable alternative for companies seeking AI independence.