Former OpenAI CTO Mira Murati Launches Thinking Machines' Inkling: Open-Weights Multimodal AI Model with 975 Billion Parameters and Self-Training Demo
Releasing yet another powerful language model is no longer enough to stand out. Thinking Machines, the startup founded by former OpenAI technical director Mira Murati, has chosen a different approach to attract developers. The company unveiled Inkling, its first open-weights model that can be freely used and adapted for specific tasks.
Inkling is a multimodal model capable of processing text, images, and audio without separate processing modules for each modality. It is built on a mixture-of-experts architecture containing 975 billion parameters, of which only 41 billion are active at any given time. This design significantly reduces computational costs. The model supports a maximum context length of 1 million tokens and was trained on a dataset of 45 trillion tokens that included text, images, audio, and video. Alongside the main model, the company released a preliminary Inkling Small version with 12 billion active parameters optimized for lower-cost and faster deployment.
One of Inkling’s standout features is the ability to regulate reasoning depth. Developers can choose how many computational resources the model should allocate to solving a problem. For simple queries the system responds faster and consumes fewer tokens, while for complex tasks it can increase internal computation. According to the company, in certain programming scenarios Inkling uses approximately three times fewer tokens than several other open models while maintaining comparable quality.
Thinking Machines acknowledges that Inkling does not yet aim to be the strongest model on the market. Closed models from OpenAI, Anthropic, and Google continue to lead most comprehensive benchmarks, while Chinese models remain ahead in certain disciplines. Instead of competing for top rankings, the company focused on creating a versatile foundation for subsequent training and customization for enterprise use cases.
To support further development, Thinking Machines offers its own Tinker platform. Through this platform, developers can fine-tune Inkling on their own data without building complex infrastructure themselves. To demonstrate the platform’s capabilities, the company conducted an unusual experiment: Inkling was tasked with training itself. The model autonomously created a training task, executed the process via Tinker, evaluated the results, and switched to the updated version. In the demonstration, it was given the unusual objective of learning to answer questions while completely avoiding the use of one letter of the English alphabet.
The company has also made the model weights available on Hugging Face and added support for popular inference frameworks including Transformers, vLLM, SGLang, and llama.cpp. Inkling is designed not only for cloud services but also for continued training, creation of specialized assistants, and development of autonomous agents capable of executing complex sequences of actions.
The release of Inkling represents an important development for the Western open-weights AI community. After Meta reduced its activity in this area and many organizations began turning to Chinese models, developers now have another major Western-origin solution that can be freely run, modified, and adapted to their own tasks thanks to its open weights.
Related articles
Why Russian Educators Prepare Sixth Graders for Ninth-Grade All-Russian Olympiads in Robotics and Information Security
Maxim Ivankov, who has run robotics and programming schools for children for nine years in a small Russian town, explains the decision to train students starting from fifth grade for the All-Russian Olympiad for Schoolchildren in the ninth-grade category. The strategy addresses the fact that regional and final stages begin only at ninth grade, while earlier participation yields only certificates with no benefits such as BVI university admission or regional prizes. The approach gives students up to five attempts instead of one, allowing systematic skill building similar to consistent athletic training. Challenges include extremely low enrollment despite free classes and widespread use of neural networks to solve school-stage tasks remotely, which distorts results and lacks regulatory prohibitions. The 2025/26 season introduced four separate profiles under informatics, including information security and robotics, both of which grant the same BVI privileges as mathematics or physics. Detailed analysis of past tasks shows a steep difficulty jump from municipal to regional stages, with topics such as the RANSAC algorithm and ten-dimensional hypercube traversal appearing only at the final level.
Corporate Boomerang: WordPress Founder Matt Mullenweg Survives Board Coup at Automattic
Automattic experienced a rapid corporate power struggle when its board placed founder and CEO Matt Mullenweg on forced paid leave. Within two days Mullenweg regained control by leveraging his shareholder voting rights and removed opposing administrators from the company Slack. The original board members who supported his ouster are now departing the company. The brief conflict highlighted how corporate governance rules allowed shareholders to ultimately determine board composition. Mullenweg described his restored position as that of a pirate before announcing renewed alignment with the board. Public support from WordPress leadership helped solidify his return to leadership.
Russian Transport Ministry Ready to Provide Regulatory Framework for Poplar Fluff Vacuum Robots
The Russian Ministry of Transport has signaled readiness to develop necessary regulations if municipal services express interest in specialized robots designed to collect poplar fluff from city streets. The statement from Minister Andrey Nikitin follows a viral meme that originated from a two-year-old video created by the agency Out Digital. Although no real machines called пухосос exist and no serial production has been launched, the fictional concept gained widespread attention after being presented as an actual Moscow development. The Department of Housing and Utilities in Moscow playfully expanded the joke by introducing additional fictional devices named Zasosyor, Musorozhor, and Gazonyukh. Linguistic experts at Gramota.ru later analyzed the newly coined term, further amplifying the meme's reach across Russian media and social networks. The minister emphasized that any future regulatory support would depend entirely on demand from communal services rather than proactive development by the ministry itself.
Context is Everything: How to Avoid Drowning in Security Incidents and Distinguish Employees from Attackers
The article explains that false positives in security systems often stem from missing business context rather than technical flaws. It outlines three levels of filtering that combine process rules, access logs, and job-specific behavior to rank risk accurately. Behavioral analysis tools like UEBA can help but require proper training and human oversight to avoid generating more noise. Key metrics focus on reducing false alerts, improving MTTD and MTTR, and minimizing daily administration time. For smaller companies without dedicated SOC teams, the guidance emphasizes starting with log collection, identifying existential risks, and aligning policies with real business processes. The piece stresses that technology alone cannot replace analysts who deeply understand company operations.