Inkling: Mira Murati's 975B Open-Weights LLM

> **TL;DR:** Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, has released Inkling, a 975-billion-parameter open-weights language model. It's the company's first frontier-scale open model release, and it quickly became one of the most-discussed launches on Hacker News. The release signals that Thinking Machines Lab intends to compete at the very top end of open-weights AI, not just in smaller research tooling.

Key Takeaways

- Inkling is a 975-billion-parameter open-weights LLM from Thinking Machines Lab - It's the startup's first frontier-scale open model release since Mira Murati founded the company - The launch drew heavy engagement and debate on Hacker News - Open weights mean developers and researchers can download and run the model themselves, unlike closed frontier models - The release puts Thinking Machines Lab squarely in competition with other labs building large open-weights systems

What Is Inkling?

Inkling is a 975-billion-parameter open-weights large language model released by Thinking Machines Lab, the AI research startup founded by Mira Murati, who previously served as OpenAI's chief technology officer. The [official announcement](https://thinkingmachines.ai/news/introducing-inkling/) and accompanying [model page](https://thinkingmachines.ai/inkling/) mark the company's first release at true frontier scale — a step up from the smaller-footprint tools and research previews it had shared since its founding.

At 975 billion parameters, Inkling sits comfortably in the size class of today's largest publicly available models. That scale alone is notable: training and serving a model this large requires substantial compute and engineering investment, and choosing to release it with open weights rather than keeping it behind an API is a deliberate strategic move rather than a routine product update.


Why Open Weights, Not Just an API

The most consequential detail in this release isn't the parameter count — it's the distribution model. Open weights mean the trained parameters of Inkling are made available for researchers, developers, and companies to download and run on their own infrastructure, rather than being locked behind a hosted API the way most frontier models from OpenAI, Anthropic, and Google are. That distinction matters for anyone tracking the broader [LLM landscape](https://speka.info/llm-updates/): it gives independent developers, startups, and enterprises the ability to fine-tune, audit, and deploy the model on their own terms.

For a company founded by someone who spent years at the center of closed-model development, releasing a 975B model with open weights is a pointed signal. It positions Thinking Machines Lab less as a challenger trying to out-ship OpenAI on closed, hosted intelligence, and more as a lab betting that openness — and the ecosystem effects that come with it — is the more durable strategy at frontier scale.

The Hacker News Reaction

Inkling's release generated substantial discussion on [Hacker News](https://news.ycombinator.com/item?id=48924929), a venue where technical audiences tend to scrutinize claims closely rather than take launch announcements at face value. That level of engagement on its own is a useful signal: frontier-scale open-weights releases are rare enough that each one draws intense interest from engineers who want to actually pull the weights, benchmark them, and see how the model behaves outside of a vendor's marketing.

The attention also reflects how closely the industry is watching Thinking Machines Lab specifically. Since Murati's departure from OpenAI, every move from her new company has been read as a data point on where former OpenAI leadership thinks the field should go next — and a first frontier-scale open release answers that question more concretely than any interview could.

What This Means for the Open-Weights Race

Inkling's arrival adds another serious entrant to a field that has been getting more crowded and more competitive at the high end. Large open-weights releases used to be dominated by a handful of players; now, a startup barely past its founding stage has shipped a model at a scale that would have been considered exclusively frontier-lab territory just a couple of years ago. That compresses the gap between what's available behind closed APIs and what any developer can download and self-host.

It's also a reminder that the broader AI ecosystem is diversifying fast on multiple fronts at once — new foundation models, new creator economics like the one built around [ElevenLabs' voice marketplace](https://speka.info/blog/elevenlabs-voice-marketplace-how-creators-earn-royalties), and increasingly specialized tooling such as [Juggler, the JUCE creator's GUI coding agent](https://speka.info/blog/juggler-juce-creators-new-gui-coding-agent). Meanwhile, enterprise adoption keeps shifting too — Microsoft recently made [GPT-5.6 the preferred model inside 365 Copilot](https://speka.info/blog/gpt-5-6-is-now-microsoft-365-copilots-preferred-model), underscoring how much competitive movement is happening across both closed and open ends of the model spectrum simultaneously.

For now, the concrete facts are these: Thinking Machines Lab has shipped a 975-billion-parameter open-weights model, its first at frontier scale, and the developer community has taken notice. What gets built on top of Inkling — and how it stacks up once independent benchmarks and real-world deployments accumulate — will be the next chapter to watch.

Frequently Asked Questions

What is Inkling?

Inkling is a 975-billion-parameter open-weights large language model released by Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati.

Who founded Thinking Machines Lab?

Thinking Machines Lab was founded by Mira Murati, who previously served as OpenAI's chief technology officer before starting the company.

Is Inkling the first model from Thinking Machines Lab?

Inkling is described as the startup's first release at frontier scale, marking a step up to a full large-scale open-weights model.

What does 'open-weights' mean for Inkling?

It means the model's trained parameters are publicly available to download and run, rather than being accessible only through a closed, hosted API.

How did the AI community react to Inkling's launch?

The release drew significant discussion and engagement on Hacker News, reflecting strong technical interest in a frontier-scale open model from Murati's startup.

Sources

- https://thinkingmachines.ai/news/introducing-inkling/ - https://thinkingmachines.ai/inkling/ - https://news.ycombinator.com/item?id=48924929