Liquid AI launches open-weight D1 decision models
inference multimodal
| Source: Mastodon | Original article
Liquid AI has launched two open-weight decision models, the d1-3B and d1-omni-600M, featuring multimodal inputs and fast inference.
Liquid AI announced the launch of two open‑weight decision models – the d1‑3B and the d1‑omni‑600M – expanding the company’s portfolio of multimodal AI tools that return structured outcomes rather than streams of generated text. Both models accept text, image and audio inputs and deliver calibrated probabilities across a predefined set of choices in a single inference call, eliminating token‑by‑token generation. According to the company’s documentation, the larger d1‑3B matches or outperforms the GPT‑6.1 Sol benchmark on four of six real‑world tasks, while the smaller d1‑omni‑600M is designed for edge devices, running in milliseconds on platforms ranging from NVIDIA DGX servers to Jetson modules.
The release matters because decision models represent a distinct class of AI that is purpose‑built for structured decision‑making, a niche that has been gaining traction in recent months. Earlier this week we reported on AutoTrust AI’s open decision model, which highlighted the speed and reasoning benefits of this approach. Liquid AI’s open‑weight offering pushes the concept further by providing the weights publicly, enabling researchers and developers to fine‑tune or embed the models directly into proprietary pipelines without relying on a hosted service. The multimodal capability also widens the range of applications—from visual inspection and audio‑driven alerts to text‑based risk assessment—while the sub‑second latency makes the models viable for real‑time edge deployments such as autonomous robotics, industrial monitoring and low‑latency recommendation systems.
What to watch next is how quickly the community adopts the d1 series and whether third‑party benchmarks confirm the claimed performance edge over leading language models. Liquid AI has already opened an API with Python and TypeScript SDKs, so integration into existing workflows should be straightforward. Follow‑up announcements may reveal larger variants, additional modality support or collaborations with enterprise partners seeking on‑premise decision‑model solutions. The open‑weight stance also invites contributions that could accelerate innovation in calibrated, token‑free AI decision making across the Nordic tech ecosystem and beyond.
Sources
Back to AIPULSEN