Mira Murati, former OpenAI CTO and now CEO of Thinking Machines Lab, Drops Its First Model Inkling, a 975-bil
- Mira Murati, the former OpenAI CTO and now a leader at Thinking Machines Lab, helped launch the company’s first model, Inkling [wired][techcrunch].
- Inkling is a 975‑billion‑parameter, open‑weight, multimodal model trained from scratch to understand text, audio, images and video; it’s reported to perform well across many tasks (reasoning, coding) though not top on every benchmark [wired][techcrunch].
- The release is positioned to help Thinking Machines compete with majors like OpenAI and Anthropic by offering a large, open‑weight alternative that requires specialized compute (cluster of chips) to run [wired][techcrunch].
Follow-up Questions:
1. How does Inkling’s architecture (e.g., mixture‑of‑experts) compare to OpenAI’s models?
2. What license covers Inkling and what limits (if any) does it place on use?
3. What hardware and cost are required to run Inkling in practice?
4. How did Thinking Machines train Inkling (data sources and compute scale)?
5. What safeguards or safety measures did the company include with Inkling?
Sources
- Thinking Machines Lab Drops Its First Model | WIRED
- Open-weight AI Model Inkling Launches by Thinking Machines
- Murati’s Thinking Machines releases first AI model for broad use | Fortune
- Thinking Machines Lab Releases Inkling Open-Weights Multimodal Model Under Apache 2.0 License: 12 Sources (Western Mainstream: 3) | NewsCord | NewsCord
- Thinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling | TechCrunch
Related questions
- How does Inkling’s architecture (e.g., mixture‑of‑experts) compare to OpenAI’s models?
- What license covers Inkling and what limits (if any) does it place on use?
- What hardware and cost are required to run Inkling in practice?
- How did Thinking Machines train Inkling (data sources and compute scale)?
- What safeguards or safety measures did the company include with Inkling?