- Outputs probabilities instead of natural language, making it easy to wire into business logic
- Answers up to 400 questions per run, covering gestures, emotions and other judgments
- Distributed via AI/ML API at about $0.0027 per demo run, targeting low-cost practical use
AI CURATED HOT STORY · 11/21/2026
HeyGen Launches Decision 1 Multimodal Probability Model on OpenRouter
HeyGen has launched Decision 1, a multimodal probability model now available on OpenRouter. Unlike generative models, Decision 1 does not produce natural language; instead, it accepts video, audio, image, and text inputs and returns calibrated probability scores for each question, enabling direct integration into business logic. A single call can handle up to 400 questions with approximately 100 ms latency. Input pricing is 50% off until December 1, with Turbo at $0.042 per million tokens and Decision 1 at $0.07, while output is free.
According to multiple sources, Decision 1 excels at batch probability judgments. In an official demo, it correctly answered all 8 questions on gesture and emotion recognition at a cost of about $0.0027 per call, highlighting its low-cost utility. The model is distributed through the AI/ML API platform and OpenRouter, offering developers flexible access. Its probability-based output is designed for seamless embedding into automated workflows such as identity verification, liveness detection, and content moderation.
The launch signals a shift from generative AI to decision-oriented AI, providing quantifiable probability tools for video review, financial authentication, and interactive entertainment. With aggressive pricing and batch processing, HeyGen aims to lower the barrier for multimodal judgment, positioning probability-based AI services as a new infrastructure layer.
Hourly heat curve
13 fully observed hoursThe line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.
Coverage timeline
Total 2 reports · Latest first- Rohan Paul (@rohanpaul_ai)T2·58 pts
- OpenRouter (@OpenRouter)T1·62 pts
- A single request accepts video, audio, image and text and returns a calibrated probability per question
- Up to 400 questions per call with roughly 100 ms latency
- Input pricing is 50% off through December 1: $0.042 per 1M tokens for Turbo and $0.07 for Decision 1, with free output