d1-3B
Liquid’s open 3B decision model, fast enough for robots and edge devices.
d1-3B is an open-weight Liquid AI decision model built on LFM2.5-VL-3B. It answers typed questions about text and images in a single forward pass instead of generating tokens. Liquid reports 48.57 on Decision Index v0.2.1 (best under 10B) and 8 ms per decision on an RTX 4090 (vendor-reported).
- Provider
- Liquid AI
- Type
- Decision model
- Released
- Oct 7, 2026
- Input
- text, image
- Output
- decision
- Open weights
- Yes
Best for
- Extraction & classification
- Real-time / low latency
- Local & on-device
- Vision
Strengths
- Runs on Jetson-class hardware
- Image input
- Open weights
More from Liquid AI
- LFM2.5-2.6B — Liquid AI. Compact on-device reasoning model on Liquid’s hybrid architecture.
- LFM2.5-24B-A2B — Liquid AI. Liquid’s largest LFM: a 24B MoE reasoning model with about 2.45B active parameters.
- LFM2.5-Embedding-350M — Liquid AI. Tiny open embedding model for on-device semantic search.