Mira Murati launches first open AI model from Thinking Machines Lab: Inkling promises to revolutionize multimodal processing

Thinking Machines Lab, a startup founded by Mira Murati, former OpenAI CTO, officially launched Inkling, its first open-weight artificial intelligence model. The model was trained from scratch to process multimodally audio, video and text, allowing developers and startups to download, modify and integrate technology into their applications without relying on closed APIs.

What is the open AI model Inkling and why is it important?

Inkling is an open-weight model, which means that any developer can download trained weights and adapt them to specific needs. Unlike black-box models such as GPT-4, Inkling allows startups and researchers to perform fine-tuning without relying on external APIs. This is particularly relevant at a time when transparency and autonomy are increasingly valued in the AI ecosystem. According to the publication of Zamin.uz, the release represents a milestone for the Thinking Machines Lab, which seeks to democratize access to advanced multimodal models.

How has Inkling open AI model been trained for audio, video and text?

According to startup, the training was performed from scratch, with a native multimodal architecture. Instead of docking separate modules for each type of data, the model is designed to learn joint representations of audio, video and text. This allows him to understand correlations between speech, images and text in a cohesive way, paving the way for applications such as video analysis with automatic transcription, multimodal search and more contextual virtual assistants. The native approach should reduce latency and improve accuracy in tasks requiring integration of multiple sources of information.

Who can benefit from Inkling Open AI model?

Inkling's open-weight approach is especially attractive for independent startups and developers who want to build AI products without the high licensing costs of proprietary models. In addition, academic researchers can use the model for experiments and innovations. Brazilian companies operating with media processing, such as edtechs and streaming platforms, can explore Inkling to create localized solutions. Meanwhile, big techs like Meta follow different paths — the company recently launched Muse Spark: the AI model that promises personal super intelligence, a closed model that contrasts with Inkling's open philosophy.

Frequently Asked Questions about the Open AI Model Inkling

Is Inkling totally free?

The model is open-weight, but there may be commercial usage restrictions depending on the license adopted by Thinking Machines Lab.

What languages does Inkling support?

Official information does not detail specific languages, but because it is multimodal, multiple language support is expected, including Portuguese.

Can Inkling run on local hardware?

Yes, because it is open-weight, the model can be downloaded and run on local servers, provided it has adequate computational capacity.