d1-3B
d1-3B is an open-weight decision model that produces an answer in a single forward pass rather than generating tokens. It accepts text and image inputs and is available on Hugging Face for local use.
What it does
Use it for yes/no decisions, choosing among named options, or scoring tasks such as routing, moderation, extraction checks, reranking, and visual inspection.
- Availability
- Available on Hugging Face; demos are available in a Hugging Face Space.
- Pricing model
- Not established
- Deployment
- Local inference on GPUs, Apple M5 Pro, and NVIDIA Jetson devices is documented.
- Privacy
- Not established
- Licence
- Open-weight; download, fine-tune, and deploy without restrictions
Capability
Modality
Access
Deployment
Pricing
Target user
Evidence
7 claims- d1-3B is available on Hugging Face. [1]
d1-3B and d1-omni-600M models are available today on Hugging Face.
Liquid AI. Quoted text is outside CC BY 4.0 and keeps its original rights. - d1-3B has day-one support for llama.cpp. [1]
d1-3B and d1-omni-600M run the full NVIDIA stack — from DGX in the data center, to RTX workstations, to Jetson at the edge — with day-one support for llama.cpp.
Liquid AI. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model can be loaded from Hugging Face as LiquidAI/d1-3B using Transformers. [2]
- d1-3B returns calibrated, typed answers in one forward pass with zero output tokens. [2]
typed answers in **one forward pass with zero output tokens**.
Liquid AI. Quoted text is outside CC BY 4.0 and keeps its original rights. - Liquid AI recommends it for yes/no decisions, named-option selection, or ratings, and says it is not a chat model and does not write text. [2]
It is not a chat model and does not write text.
Liquid AI. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model has 3.12B total parameters and a context length of 32,768 tokens. [2]
- It accepts text, JSON, images, or a mixture as its state. [2]
You give it a state (text, JSON, images, or a mix) and a set of questions.
Liquid AI. Quoted text is outside CC BY 4.0 and keeps its original rights.
Sources
3
1
Verified
2
Verified
3
Verified
Versions
2| Version | Effective | Change | Changed fields | State |
|---|---|---|---|---|
| 2 | 2026-10-08 20:34 UTC | material change | capability, modality, access, deployment, availability, deployment | Current |
| 1 | 2026-10-07 21:26 UTC | publication | None | Current |
In briefs
1- Liquid AI’s decision models classify inputs without writing a response Daily brief 2026-10-07
Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.