Liquid D1-3B
3B decision model that answers in milliseconds instead of generating tokens.
About
A different kind of model: instead of generating text, d1-3B returns calibrated yes/no, choice, and score answers in a single forward pass — 8ms on an RTX 4090, 16ms on a Jetson AGX Thor. It takes text and images, scores 48.57 on Liquid's Decision Index 0.2.1 (best under 10B, per the company), and is meant for structured choices inside larger systems like routing, moderation, and approvals.
The catch: it's a decision head, not a conversational model, and deployment needs transformers 5.14+ with trust_remote_code.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- decision-index-0-2-1: 48.57 (Liquid-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: Released Oct 7, 2026 — very new; two open-weight decision models (d1-3B, d1-omni-600M); Decision model, not conversational: single forward pass, zero output tokens, typed answers; Decision Index 48.57 and latency figures are Liquid-reported; deployment needs transformers>=5.14 with trust_remote_code. License: License terms not stated in the sources read; confirm on the model card before commercial use..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 5.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · Released Oct 7, 2026 — very new; two open-weight decision models (d1-3B, d1-omni-600M)
- · Decision model, not conversational: single forward pass, zero output tokens, typed answers
- · Decision Index 48.57 and latency figures are Liquid-reported; deployment needs transformers>=5.14 with trust_remote_code