Lower-cost sibling of Cohere's Embed 5 Pro, for high-volume multimodal embedding and indexing across 100+ languages
d1.2
New· 4Liquid AI's faster hosted decision model, released 16 November 2026: returns calibrated probabilities and typed answers from text, images or audio in one call, at $0.02 per million input tokens and no output charge.
TrustList summary
Not yet independently verified. Liquid AI's own blog post of 16 November 2026 and its decision-model documentation supply the model codes, prices, input types and every performance figure, and the figures are the vendor's own, which we have not verified. The only independent source we found is OpenRouter's listing of the model, dated 16 November 2026, which shows the same input price and no output charge. We found no independent test, and we did not open the Artificial Analysis index the vendor cites. The vendor states no context window or licence for the hosted model, so neither is recorded here. We will update this when it can be confirmed, and remove this note.
Most language models answer a question by writing text. Liquid AI's d1.2 answers by choosing: it reads the situation and returns a calibrated probability for each possible outcome in a single call, with no generated tokens. The vendor released it on 16 November 2026 together with a larger sibling, d1.2-ultra, as the new versions of its decision models, and it is the cheaper and quicker of the pair, listed by Liquid AI as less than half the price of the ultra model.
Five new kinds of answer
The release adds five tasks to the Liquid API. A model can read a value out of the input as text, a number or a date, and says "not found" instead of guessing when the document does not show it. It can find every passage of a kind in a text, such as each piece of personal data in a note, with its position. It can locate things in a picture, returning one box or a box for every instance together with a count. It can estimate a number and give a range, and it can pick every option that applies, each with its own probability. The new answer types use the same endpoint and request format as before.
Audio as input
Audio can now be what the model is asked about, for example which language someone speaks or whether a siren sounds in the background, and the question itself can be a recording instead of typed text. Liquid AI describes the model as omnimodal because text, images and audio all go in through the same request.
What the vendor claims
Liquid AI says that no model scores higher than d1.2 for less money on the Artificial Analysis Decision Index v2.1, and that d1.2 has a calibration error of 1.2 points across 13 benchmarks and 22,505 questions. For the new tasks it reports its own tests on held-out splits of public datasets, for instance 96% on reading values from receipts. These are the vendor's figures, and no independent test has been published.
Billing for a request
Input tokens are the only charge. A request pays for its text, pictures and audio once, and each extra question adds just its own text. A 1024 by 1024 image counts as 1,536 tokens. OpenRouter lists the model at the same $0.02 per million input tokens from 16 November.
Key facts
Maker: Liquid AI Released: 16 November 2026 Model code: d1.2 Inputs: text, images, audio Price per million tokens: $0.02 input, no output charge Access: Liquid API, with a key from the Liquid console; also listed on OpenRouter Licence: not stated for the hosted model
Benchmarks & AI stats
| Benchmark | Official | Community 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.
Trust Score
An earned signal from verification, reviews, awards, transparency and engagement - the vendor can't buy it.
Updated 11/17/2026
Request a demo or quote from d1.2
Reviews
Questions & answers
No questions yet. Be the first to ask about d1.2.
d1.2 alternatives
Similar listings buyers compare against this one.
Open-weight translation model from Bilibili's Index team, covering 150 languages and following glossary and style instructions
Cohere's larger multimodal embedding model for enterprise search and retrieval, handling text, images and documents in 100+ languages
Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection
Low-latency version of ElevenLabs' Eleven v4 speech model, aimed at phone and chat voice agents in 90+ languages
Aleph Alpha's open-weight English-German mixture-of-experts model for self-hosting, released under Apache 2.0 in October 2026
Microsoft AI's high-volume, low-latency variant of MAI-Voice-2.1 for speech in 23 languages, aimed at voice agents
Microsoft AI's text-to-speech model for keeping one voice across 23 languages, with voice cloning from short audio
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- Write a reviewHelp other buyers decide
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.