Liquid AI Releases d1-3B and d1-omni-600M Models
Liquid AI launches two open-weight multimodal decision models, d1-3B and d1-omni-600M, featuring zero output tokens.
Stock photo for illustration only, not from the actual event
- Liquid AI introduces Open d1, featuring two new open-weight decision models
- The d1-3B model processes text and image inputs
- The d1-omni-600M model supports text with images or text with audio
- Neither model generates or writes any text outputs
Liquid AI has officially released Open d1, introducing two new open-weight multimodal models within its d1 decision model family, namely d1-3B and d1-omni-600M, specifically engineered for rapid decision-making tasks.
The d1-3B model is capable of reading and interpreting both text and images, whereas the d1-omni-600M model provides versatile input options by processing text alongside images or text combined with audio data.
A distinctive characteristic of these models is their inability to write text. Instead of generating traditional text responses, each model returns calibrated and typed answers in a single forward pass with zero output tokens.
Stock photo for illustration only, not from the actual event
The introduction of Liquid AI's d1 models highlights a significant paradigm shift in AI design. Rather than focusing on generative text capabilities like standard large language models, these architectures prioritize deterministic and rapid decision outputs, making them ideal for low-latency automated environments.
The overarching objective for these multimodal decision models is to facilitate real-time decision-making processes across various applications by eliminating unnecessary generative overhead and streamlining data processing.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment