Liquid AI’s D1 Adds Low-Cost Decision Inference to OpenRouter
D1 handles structured classification, routing, ranking, and risk checks with probability estimates at low input cost.
TL;DR:
- D1 returns choices, scores, and probabilities instead of generating text.
- Low input pricing fits high-volume classification, routing, ranking, and risk checks.
- Probability outputs let teams set confidence thresholds and send uncertain cases to humans or other models.
- OpenRouter adds a decision-specific model; Liquid AI reaches more developers.
- Success hinges on whether D1 beats cheap classifiers or larger models in accuracy and calibration.
Headline: Liquid AI’s D1 offers low-cost probabilistic decision inference on OpenRouter.
Summary: OpenRouter now hosts Liquid AI’s D1 for decision tasks. Input app state to get yes/no, choice or score outputs with probabilities. It supports 65K token context, keeps data private, charges $0.04 per million input tokens and nothing for output. It suits high-volume classification, routing, ranking and risk checks.
Analysis: D1 focuses on reliable structured outputs and accurate probability estimates instead of generating text. The low price may encourage teams to use it for everyday decisions in place of rule-based systems or larger models. The probabilities allow setting thresholds, routing low-confidence cases to humans. OpenRouter gains a tool for specialized routing. Liquid AI gains access to developers who mix models. It remains to be seen whether it outperforms cheap classifiers or bigger models on accuracy.
Impact Assessment:
- Significance: Medium
- Categories: Model Release, Developer Tools, Industry Trend