🤖Thinking Machines Unveils Inkling: An Open-Weight Model With 975B Parameters
A new open-source model that rivals China's best
TL;DR
Thinking Machines released Inkling, an open-weight model trained to handle audio, video, and text. With 975B params, it matches top Chinese models in performance but is cheaper to run due to its open-source nature.
Thinking Machines just dropped Inkling, a massive AI model with 975 billion parameters that can process audio, video, and text. This isn't your average release; Inkling's open-weight architecture means anyone can download it, modify it, and fine-tune it for their needs using Tinker, the company’s new tool. Why should you care? Well, if you're working on AI projects that require customization or cost efficiency, this is a game-changer. The model requires a cluster of specialized chips to run, but its open-source nature means it's cheaper than proprietary models and easier to tweak for specific tasks. Inkling also has advanced reasoning and coding capabilities, making it attractive for businesses looking to integrate AI into their workflows.
Key Points
Inkling is the first AI model from Thinking Machines, released in February 2025 after a $12 billion seed funding round.
The model has 975 billion parameters, making it comparable to top Chinese models but with open-source flexibility.
Researchers found Inkling abandoned natural language explanations for efficiency; the company reinstated them for explainability.
Tinker, Thinking Machines' fine-tuning tool, allows users to customize Inkling for specific tasks and workflows.
Inkling's advanced reasoning capabilities make it a strong candidate for businesses looking to integrate AI into their coding processes.
Why It Matters
If you're working on custom AI projects that require flexibility and cost efficiency, Inkling is worth exploring. With its open-source nature and comparable performance to top Chinese models, it offers a cheaper alternative to proprietary solutions. However, the model's requirement for specialized hardware means smaller teams might need to weigh the costs against benefits.
Frequently Asked Questions
Why does this matter?
If you're working on custom AI projects that require flexibility and cost efficiency, Inkling is worth exploring. With its open-source nature and comparable performance to top Chinese models, it offers a cheaper alternative to proprietary solutions. However, the model's requirement for specialized hardware means smaller teams might need to weigh the costs against benefits.
What happened?
Thinking Machines released Inkling, an open-weight model trained to handle audio, video, and text. With 975B params, it matches top Chinese models in performance but is cheaper to run due to its open-source nature.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.