TL;DR
Thinking Machines, led by Mira Murati, is developing AI ‘interaction models’ that respond to users in real time across multiple modalities. The company plans a limited research preview soon, with a wider release later this year.
Thinking Machines, the AI firm founded by Mira Murati, announced it is developing ‘interaction models’ that allow AI to respond in real time across audio, video, and text, marking a significant step toward more natural human-AI collaboration.
The company describes these ‘interaction models’ as systems that process multiple modalities simultaneously, enabling AI to perceive ongoing human input and respond dynamically. Unlike current models that process input sequentially and wait for completion, these new models aim to operate continuously, capturing real-time audio, video, and text data. According to Thinking Machines, this approach addresses the ‘bandwidth bottleneck’ of traditional AI interfaces, which can hinder effective collaboration.
Mira Murati, founder of Thinking Machines, stated that the goal is to create AI systems that meet humans where they are, rather than forcing users to adapt to rigid interfaces. The company showcased early examples, including real-time speech translation, detection of mentions of animals in stories, and alerts about user posture. However, the company emphasized that these features are still in development and are not yet publicly accessible.
Thinking Machines plans to launch a ‘limited research preview’ of these interaction models in the coming months, with a broader release targeted later this year. The company is currently in the early stages of testing, and it is not yet clear how widely available the models will be or how they will be integrated into products.
Why It Matters
This development could significantly advance human-AI interaction, making AI systems more intuitive and responsive. If successful, it could impact various fields such as communication, translation, virtual assistance, and collaborative work, potentially transforming how humans engage with AI tools on a daily basis. The move also signals a shift toward more dynamic, real-time AI systems that can process complex, multi-modal data streams, which has been a longstanding challenge in AI research.
AI Translation Earbuds Real Time 164 Languages 80H Playtime Translator Ear Buds Audifonos Traductores Inglés Español Wireless Earphones Bluetooth AI Headphone for Travel Meeting Learning K08 Black
Supports 164 Languages Worldwide: Powered by cutting-edge AI translation technology, these translator earbuds real time support translation in…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
Mira Murati founded Thinking Machines in February 2025 after leaving OpenAI, aiming to push the boundaries of AI interaction. The company’s focus on real-time, multi-modal AI models comes amid broader industry efforts to improve AI responsiveness and contextual understanding. Previous developments in AI have largely centered on text-based models or delayed responses, limiting natural collaboration. Murati’s departure from OpenAI and the subsequent formation of Thinking Machines marked a strategic pivot toward more interactive AI systems. The company has faced some internal turnover, with key members leaving for other tech giants like Meta and returning to OpenAI, but it remains committed to advancing this new form of AI interaction.
“Our goal is to make AI truly interactive across any modality, so it can collaborate with humans in real time, just like we do with each other.”
— Mira Murati
“We believe that addressing the bandwidth bottleneck in current models will unlock new possibilities for AI in everyday life.”
— Thinking Machines spokesperson
Multi-Modal User Interactions in Controlled Environments (Multimedia Systems and Applications, 34)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It is not yet clear how widely available the interaction models will be, what specific applications they will support at launch, or how they will be integrated into existing AI systems. Details about the timeline for broader deployment and the potential limitations of the technology remain uncertain as the company is still in early testing phases.
AI Tools for Video Creators
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
Thinking Machines plans to roll out a limited research preview in the coming months, allowing select partners and researchers to test the interaction models. The company aims for a wider release later this year, with further details likely to emerge as testing progresses and feedback is gathered.
Abacus Brands Virtual Reality ESPN Gift Box – Illustrated Interactive VR Book and STEM Learning Activity Set
EXPERIENCE SPORTS IN VIRTUAL REALITY! Dive into the world of sports like never before with immersive experiences in…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What are ‘interaction models’ in AI?
Interaction models are AI systems designed to process and respond to multiple modalities—such as audio, video, and text—in real time, enabling more natural and dynamic collaboration with humans.
When will the public be able to try these models?
Thinking Machines plans to launch a limited research preview in the coming months, with a broader release expected later in 2026.
How do these models differ from current AI systems?
Current AI models typically process input sequentially and wait until input is complete before responding. These new interaction models aim to operate continuously across multiple modalities, providing real-time responses and more fluid interaction.
Who is Mira Murati, and what is her background?
Mira Murati is the founder of Thinking Machines and a former CTO of OpenAI, known for her leadership in AI development and her focus on advancing interactive AI systems.
What challenges does this technology aim to solve?
The models aim to overcome the ‘bandwidth bottleneck’ in traditional AI, which limits how much human knowledge and intent can be effectively communicated and understood by AI systems in real time.
