Thinking Machines Lab has released Inkling, a 975-billion-parameter open-source AI model trained to process both video and audio alongside text. The model represents the company's first public offering after 18 months of building AI infrastructure largely out of public view, positioning itself as an alternative to the dominant one-size-fits-all approaches from competitors like OpenAI and Anthropic.

Inkling's focus on multimodal capabilities—particularly video and audio understanding—suggests Thinking Machines is targeting use cases where traditional text-only language models fall short. By releasing it as open source, the company aims to attract developer adoption and establish credibility within the AI community.

What This Means for Your Business

Companies working with video or audio analysis—from content moderation to security surveillance to media processing—should evaluate Inkling as a potentially more cost-effective open alternative to commercial models. The open-source approach means no per-token costs and greater control over deployment, though teams will need the in-house expertise to manage and fine-tune the model themselves. This could be particularly valuable for enterprises with sensitive video data that they cannot send to third-party APIs.