Skip to main content

Liquid AI says d1.2-ultra ranks first on decision-model index

16 NOVEMBER 2026·2 MIN READ·4 SOURCES

Liquid AI says its d1.2-ultra model ranks first among decision models on Artificial Analysis’s Decision Index v2.1, scoring 75. The company also says its updated decision models add audio input and five new tasks.

Key takeaways · 4

  • 01

    d1.2-ultra scored 75 and ranks first among 24 models listed by Artificial Analysis.

  • 02

    Liquid AI lists API prices of $0.02 per million input tokens for d1.2 and $0.042 for d1.2-ultra.

  • 03

    Both models bill input tokens only; Liquid AI says a 1024-by-1024 image costs 1,536 tokens at the text rate.

  • 04

    Liquid AI says no public benchmark yet measures the five new tasks, which it evaluated using internal sets made from held-out public-dataset splits.

A new top-ranked decision model

Liquid AI’s updated decision models add audio input and five new tasks, according to the company.[1] Its d1.2-ultra model scored 75 on Artificial Analysis’s Decision Index v2.1; Liquid AI says that is two points above the previous best decision model.[1] Artificial Analysis lists d1.2-ultra at number one of 24 models, also with a score of 75.[2] The index median is 66, according to Artificial Analysis.[2] An AI Tracker report had placed d1.2—not d1.2-ultra—sixth on the index.[3]

What the models are designed to do

Liquid AI describes the models as handling tasks that include extracting a value from an invoice and returning the characters read, or answering “not found” when the invoice does not show the value.[1] They can also identify each requested kind of passage in text and return its position.[1] For images, Liquid AI says they can mark every tree with a rectangle and provide a count, or estimate a count with a range alongside a best guess.[1] Another described task selects all applicable options and assigns a probability to each selected topic.[1] The question itself can be spoken in a recording rather than typed.[1]

Index score and evaluation limits

Artificial Analysis says its composite index assesses decision models across document decisions and extraction, tool calling, knowledge, reasoning and invariance.[4] Liquid AI describes the index as covering six benchmarks, including document decisions and extraction, tool calling and hard knowledge questions.[1] The Decision Score rewards correct answers and calibrated confidence, while applying greater penalties to confident wrong answers, according to Artificial Analysis.[4] Liquid AI reports d1.2-ultra’s calibration error at 0.8 points, compared with 1.2 for d1.2 and higher figures for the other models it names.[1] The company says no public benchmark yet measures the five new tasks, so it used internal test sets made from held-out splits of public datasets.[1]

API pricing and availability

Liquid AI says both models were available on its API at launch.[1] It lists d1.2 at $0.02 and d1.2-ultra at $0.042 per million input tokens.[1] The company says both models bill input tokens only, and that a 1024-by-1024 image costs 1,536 tokens at the text rate.[1] Liquid AI reports that both models answer 95% to 97% of questions about sounds or spoken language, with language tests spanning 43 languages.[1] The company said the models were expected to come to OpenRouter and Vercel soon.[1]

Teams considering decision models can compare a ranked benchmark result with the capabilities, pricing and input-token billing Liquid AI reports. The new tasks lack a public benchmark, so organizations evaluating them should distinguish the company’s internal results from the independently conducted index evaluations.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 16 November 2026

    Liquid AI says d1.2-ultra ranks first on decision-model index

Sources

AI fluency, one session a day, built for your work.