Skip to content
TrustList
D3
AI Model

d1-3B

New· 4

Liquid AI's open-weight decision model, giving yes or no, choice or score answers from text and images

About d1-3B

Liquid AI released d1-3B on 7 October 2026 as the larger of the two open-weight models in its d1 decision family. It answers rather than writes: instead of producing text, it returns, in one pass, a calibrated yes or no, an option picked from a list, or a numeric score.

Built on a vision-language base

Its base is LFM2.5-VL-3B, a vision-language model from the same company. Parameters number about 3.12 billion and the context length is 32,768 tokens, and it accepts both text and images, so a decision can depend on a picture as well as on words.

Intended uses

Liquid AI names a long list of intended uses: sending requests to the right place, moderating content, classifying intent and topic, checking extractions, reranking results, scoring as an LLM judge, guarding agents and inspecting images. That is quite a wide list for one small model. All of these are tasks where a short, trustworthy verdict is worth more than a paragraph of generated prose. It is not a chat model.

Weights and licensing

The weights sit on Hugging Face under the licence tagged lfm1.0, and they can be downloaded, fine-tuned and deployed in a company's own systems. No price is stated.

Reported figures

Liquid AI reports a mean score of 82.9 over seven public text benchmarks, which it ran in house. A second figure, 48.57 on the Decision Index v0.2.1 public split, comes with no word on who measured it, and it published latency figures on devices from Apple, NVIDIA and AMD. These are vendor figures, so a buyer should confirm them on their own hardware.

TrustList summary

d1-3B is the bigger of two open-weight decision models that Liquid AI put out on 7 October 2026 as its d1 family. It is based on Liquid AI's LFM2.5-VL-3B vision-language model and takes text and images as input. It is not a chat model, because it returns a single-pass answer.

Main features

  • Decision answers: calibrated yes or no answers, a choice from a list, or a score, returned in one pass.
  • Routing and triage: routing, triage, moderation, intent and topic classification.
  • Agent guardrails: checks for agent systems, along with LLM-as-a-judge scoring and reranking.
  • Visual inspection: image input for visual inspection work.

Who it works with

Liquid AI lists these uses, which fit AI systems that need checks or routing rather than free text.

Key facts

  • Released: 7 October 2026
  • Parameters: about 3.12 billion
  • Context length: 32,768 tokens
  • Licence: lfm1.0, as tagged on Hugging Face

Summary written by TrustList from the business's own published information, Nov 2026.

Benchmarks & AI stats

BenchmarkOfficialCommunity avg
No benchmark scores yet. Be the first to add one.

“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).

Sign in to add a benchmark score for this model.

d1-3B launches

Trust Score

4/ 100
Trust Score: New

An earned signal from verification, reviews, awards, transparency and engagement - the vendor can't buy it.

Verification
0/100 · 20%
Reviews
0/100 · 30%
Awards
0/100 · 15%
Transparency
18/100 · 20%
Engagement
0/100 · 15%
Joining soon
Recommendations
Coming soon
Complaints
Coming soon

Updated 10/9/2026

What the business tells TrustList

Founded
2026 (0 years) evidence sent

Request a demo or quote from d1-3B

Protected by reCAPTCHA - Google Privacy Policy and Terms apply.

By sending, you agree we may share your request and contact details with the provider once you confirm your email.

Reviews

Write the first review of d1-3B

Used it? Your experience helps other buyers decide.

Write a review

Questions & answers

No questions yet. Be the first to ask about d1-3B.

d1-3B alternatives

Similar listings buyers compare against this one.

Cantina Security's first open-weight security model, a post-trained GLM-5.3-Flash for vulnerability research and exploit checks

Google's open multimodal embedding model that maps text, images, audio and video into one vector space, under Apache 2.0

Perplexity's open-weight decision model that picks among given options for routing and checks in AI systems

Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection

JetBrains' open 12B mixture-of-experts coding model with 2.5B parameters active per token under Apache 2.0

Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure

Ai2's open-weight 8B model that writes cited scientific reports from a research question and retrieved literature

Microsoft AI's high-volume, low-latency variant of MAI-Voice-2.1 for speech in 23 languages, aimed at voice agents