We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Hacker News·20d ago·5 min read
Murati's Thinking Machines Releases Open-Weights 975B Parameter LLM
TL;DR
Murati's Thinking Machines has released a 975 billion parameter language model (LLM) with open weights, allowing for extensive customization. This generalist model efficiently handles text, images, and audio, making it versatile for various applications.
✦ Why It Matters
Engineers can start experimenting with the open weights of this LLM today to create tailored AI solutions.
Key Takeaways
- Experiment with the open-weight model for diverse applications.
- Utilize the model's multimodal capabilities for innovative projects.
- Adjust thinking time to optimize performance for specific tasks.
- Leverage the model's coding abilities to automate workflows.
Full Summary
Murati's Thinking Machines has unveiled a groundbreaking open-weight large language model (LLM) with 975 billion parameters, of which 41 billion are active. This model employs a Mixture of Experts architecture, allowing it to efficiently handle diverse input modalities, including text, images, and audio.
With a context window of up to 256,000 tokens, it can process extensive information at once. The model is designed for general intelligence, performing well in knowledge, mathematics, and science tasks.
It also features capabilities for agentic coding, enabling it to write code and utilize tools effectively. Additionally, users can adjust the model's thinking time to balance speed and performance, while its forecasting abilities provide well-calibrated confidence in predictions.