Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
Data Science
Simon Willison's Blog·21d ago·5 min read
Jev introduces a new shape of LLM - System One, aka Decision Models
TL;DR
TypeSafe AI has launched Jev, a new type of model called System One or Decision Models, which outputs probabilistic decisions instead of text. Jev is cost-effective and fast, charging only for input tokens while providing free output.
✦ Why It Matters
Consider integrating Jev into your applications for cost-effective decision-making capabilities.
Key Takeaways
- Explore using Jev for decision-making tasks in your projects.
- Utilize Jev's API for processing semi-structured data efficiently.
- Take advantage of Jev's low input cost for large-scale applications.
- Experiment with Jev's output for various classification and rating tasks.
Full Summary
TypeSafe AI has introduced Jev, a novel model categorized as a System One or Decision Model, which diverges from traditional large language models (LLMs) by returning floating point numbers instead of text. This model accepts text or semi-structured data as input and generates outputs that include category classifications, yes/no answers, ratings, and confidence scores.
Jev operates by taking a 'state' object—composed of strings or name-value pairs—and processing it through an API to answer specific questions. Notably, Jev is designed to be both fast and economical, charging only $0.042 per million input tokens, which is cheaper than OpenAI's GPT-5 Nano.
This pricing model allows users to leverage Jev for various applications without incurring high costs for output. The implications of this model suggest a shift in how decision-making processes can be automated using AI.