What Is Thinking Machines Labs? Mira Murati's New AI Company Explained
Thinking Machines Labs is Mira Murati's post-OpenAI AI startup. Learn what makes their interaction model different and why AI builders should pay attention.
Who Is Mira Murati and Why Her New Company Matters
Mira Murati spent six years building some of the most influential AI systems in existence. As OpenAI’s Chief Technology Officer, she oversaw the development of GPT-4, DALL-E, and ChatGPT — products that collectively reshaped public understanding of what AI could do. She also briefly served as interim CEO during the turbulent leadership crisis of November 2023, a moment that put her squarely in the global spotlight.
In September 2024, she left OpenAI.
A few months later, she announced Thinking Machines Lab — and the AI world paid close attention. The company represents one of the most high-profile post-OpenAI ventures to emerge from the current wave of AI research, and its founding philosophy reflects a clear departure from the pure scale-and-ship approach that defined the last few years.
This article explains what Thinking Machines Lab is, what it’s trying to build, why the company’s interaction model is distinctive, and what it means for developers and AI builders working with LLMs and multi-agent systems today.
What Is Thinking Machines Lab?
Thinking Machines Lab is an AI research and product company founded by Mira Murati in late 2024. It publicly launched in early 2025 with a stated focus on building AI systems that are more interpretable, collaborative, and tightly integrated into human workflows.
Everyone else built a construction worker.
We built the contractor.
One file at a time.
UI, API, database, deploy.
The company name itself is a deliberate reference. Alan Turing’s 1950 paper “Computing Machinery and Intelligence” opened with the question: “Can machines think?” Thinking Machines Lab positions itself as working toward AI that doesn’t just produce outputs but genuinely reasons — and does so in a way humans can understand and trust.
The company raised significant early funding, drawing investment from major backers who saw the founding team’s depth of experience as a credible bet. Early reports placed the company’s valuation in the multi-billion dollar range shortly after formation, reflecting both the strength of the team and investor appetite for credible alternatives in the frontier AI space.
The Founding Team’s Background
Murati didn’t build Thinking Machines Lab alone. She brought together a core group of researchers and engineers, many with deep roots at OpenAI and other leading AI labs. That team composition matters because it signals what kind of company this is: research-first, with an eye toward novel approaches rather than simply building faster on top of existing paradigms.
The collective experience spans multimodal AI, reinforcement learning, safety, and large-scale infrastructure — a combination that maps directly to what Murati has said she wants to build.
What Makes Thinking Machines Lab Different from OpenAI?
This is the natural question, and the honest answer is: the differentiation is more philosophical than technical at this stage. But philosophy shapes product decisions, so it’s worth understanding.
Transparency Over Black-Box Outputs
One of the clearest signals from Murati’s public statements is an emphasis on interpretability — building AI that can explain its reasoning, not just produce answers. This stands in contrast to how most large language models work today, where users see output but have little visibility into why the model produced it.
Thinking Machines Lab appears to be betting that the next meaningful leap in AI isn’t just better outputs but better accountability for how those outputs are reached.
Human-AI Collaboration as a Design Principle
Murati has spoken about wanting AI to be genuinely collaborative — not a tool that replaces human thinking, but one that augments it in real-time. This sounds similar to things other labs say, but the framing at Thinking Machines Lab seems more operationally specific: building systems where humans and AI work together across long, multi-step tasks, with the human remaining meaningfully in the loop.
For multi-agent workflows specifically, this distinction matters. When AI agents are handling chains of actions autonomously, interpretability and collaboration design directly affect whether the system is trustworthy and auditable.
A Different Pace and Culture
OpenAI’s pace of product releases accelerated dramatically in the years leading up to Murati’s departure. Thinking Machines Lab, at least in its early stage, appears to be taking a more deliberate approach: fewer public announcements, more foundational research orientation, and a team small enough to maintain coherent direction.
That’s not unusual for an early-stage AI lab — but given how aggressively other frontier labs push to ship, the contrast is notable.
What Is Thinking Machines Lab Building?
The company has been deliberate about what it discloses publicly, which is consistent with labs that prioritize research momentum over marketing. That said, several focus areas have emerged clearly.
Multimodal AI Models
Other agents ship a demo. Remy ships an app.
Real backend. Real database. Real auth. Real plumbing. Remy has it all.
Thinking Machines Lab is building models that process and generate across multiple modalities — text, images, and potentially other inputs. This aligns with the trajectory Murati oversaw at OpenAI and reflects the broader industry consensus that the most capable and useful AI systems will need to operate across more than text.
AI That Reasons Transparently
A recurring theme in the company’s positioning is reasoning transparency. Rather than simply scaling transformer architectures toward better benchmark performance, Thinking Machines Lab appears interested in how models can surface their reasoning process in a way that’s meaningful to users — particularly in professional and high-stakes contexts.
This is directly relevant to the current conversation around chain-of-thought reasoning, AI verification, and the gap between a model that sounds confident and one that actually is.
Enterprise and Professional Applications
While the company hasn’t shipped a consumer product as of this writing, the focus on collaborative intelligence and interpretability points clearly toward professional use cases: legal, medical, scientific research, complex decision-making environments where outputs can’t just be plausible — they need to be auditable.
Why AI Builders and Developers Should Pay Attention
If you’re building with AI today — using LLMs, constructing multi-agent pipelines, or deploying AI-powered applications — Thinking Machines Lab is worth tracking for a few specific reasons.
A New Model Entrant in the LLM Ecosystem
The number of credible frontier model providers has grown substantially in the last two years: OpenAI, Anthropic, Google DeepMind, Meta, Mistral, and others. Thinking Machines Lab is positioned to become another serious entrant, potentially offering models with distinct properties — particularly around reasoning transparency and multimodal capability.
For builders who work across multiple models and need to select the right one for the right task, a Thinking Machines Lab model could become a meaningful option in that toolkit.
The Interaction Model Is Different
The phrase “interaction model” in this context refers to how the AI interfaces with users and how humans are positioned relative to the AI’s decision-making. Most current LLM interactions are stateless and transactional — you prompt, you receive. Multi-agent systems have begun to change that, but even there, the human-AI relationship is often hands-off once a workflow is running.
Thinking Machines Lab seems to be working toward something more continuous: AI that stays engaged with human reasoning rather than operating independently once given a task. For anyone building agents that handle complex, multi-step workflows, that design philosophy could yield models that are more suitable for supervised autonomy — capable enough to execute, interpretable enough to trust.
The Team’s Influence on Industry Norms
What companies like Thinking Machines Lab ship influences what the rest of the industry builds. When a team of this caliber invests in interpretability and collaborative design, it signals that these properties are worth competing on — which eventually shifts what customers expect and what other labs prioritize.
That matters for AI builders today because the tools and models available in two years will likely reflect the priorities being set now by labs like this one.
Thinking Machines Lab in the Context of the Broader AI Race
How Remy works. You talk. Remy ships.
The competitive landscape in frontier AI has never been more crowded or better funded. But Thinking Machines Lab isn’t competing on scale alone — which is a strategic choice, because competing on scale requires resources that only a handful of organizations can sustain.
Instead, the company appears to be making a bet that the most valuable AI systems won’t be the biggest, but the most trustworthy. That means interpretable reasoning, better calibration between confidence and accuracy, and human-AI collaboration that actually works in practice — not just in demos.
This is a long-term thesis, and it’s one that’s genuinely uncertain. It’s possible that scaling continues to solve problems faster than interpretability research. But Murati’s track record of shipping consequential AI systems — not just publishing about them — gives this lab more credibility than most.
For the multi-agent and LLM space specifically, where complexity and opacity are already problems at the system level, the Thinking Machines Lab approach could prove particularly valuable.
How MindStudio Fits Into This Picture
If you’re following Thinking Machines Lab because you’re building AI-powered workflows or agents, there’s a practical question that sits underneath the theory: how do you actually build with the best available models without locking yourself into one provider’s ecosystem?
That’s exactly the problem MindStudio is designed to solve. MindStudio is a no-code platform for building AI agents and automated workflows. It gives you access to 200+ models — including Claude, GPT-4o, Gemini, and others — from a single interface, without needing to manage separate API keys or accounts for each one.
As new model entrants like Thinking Machines Lab ship their own models, the ability to evaluate and switch models without rebuilding your entire workflow becomes more valuable. MindStudio’s model-agnostic architecture is built for exactly that kind of flexibility.
For teams building multi-agent pipelines — which is directly relevant given Thinking Machines Lab’s focus on collaborative, multi-step AI — MindStudio supports autonomous background agents, webhook-triggered workflows, and agentic MCP servers that expose your agents to other AI systems. You can build and deploy a working AI agent in an hour without writing code, and swap models as the landscape shifts.
You can try MindStudio free at mindstudio.ai.
Frequently Asked Questions
What is Thinking Machines Lab?
Thinking Machines Lab is an AI research and product company founded by Mira Murati in late 2024, following her departure from OpenAI. The company is focused on building AI systems that reason transparently, work collaboratively with humans, and are interpretable — particularly for professional and enterprise applications.
Who founded Thinking Machines Lab?
Mira Murati founded Thinking Machines Lab. She previously served as CTO at OpenAI, where she led the development of GPT-4, DALL-E, and ChatGPT. She also briefly served as OpenAI’s interim CEO during the board crisis of November 2023 before Sam Altman returned to the role.
What is Thinking Machines Lab building?
The company is developing multimodal AI models with a focus on reasoning transparency and human-AI collaboration. While specific product details remain limited, the direction points toward enterprise and professional use cases where interpretability and accountability are critical.
How is Thinking Machines Lab different from OpenAI?
The primary difference is philosophical. Thinking Machines Lab is emphasizing interpretable reasoning and collaborative human-AI interaction rather than pushing purely for scale and capability. The company is also significantly smaller and more research-oriented in its early stage, taking a more deliberate pace than OpenAI’s recent product cadence.
Has Thinking Machines Lab raised funding?
Yes. The company secured substantial early-stage funding shortly after its public launch in 2025, with reports placing its early valuation in the multi-billion dollar range. The fundraising reflected strong investor interest in the founding team’s experience and the company’s differentiated approach.
Why does Thinking Machines Lab matter for AI developers?
Thinking Machines Lab represents a potential new frontier model provider with a distinctive focus on reasoning transparency and multi-step human-AI collaboration. For developers building with LLMs or constructing multi-agent workflows, this could mean access to models better suited for supervised autonomy and auditable outputs — a meaningful addition to the growing model ecosystem.
Key Takeaways
- Thinking Machines Lab is Mira Murati’s post-OpenAI AI company, focused on interpretable, collaborative AI rather than raw scale.
- The company’s core thesis is that the most valuable AI systems will be the most trustworthy — not just the most capable.
- Their interaction model is designed for genuine human-AI collaboration across multi-step tasks, with transparency as a first-class concern.
- For builders working with LLMs and multi-agent systems, Thinking Machines Lab is worth watching as a potential new model entrant with distinct properties.
- The ability to evaluate and work with multiple models without infrastructure lock-in — as MindStudio enables — becomes more valuable as the frontier model ecosystem expands.
The AI landscape is gaining more serious players with genuinely different approaches. Thinking Machines Lab is one of the most credible. Whether you’re building agents today or evaluating what to build next, understanding what they’re working toward is a reasonable investment of your attention.
