Close Menu
NCIJ Network NCIJ Network
    What's Hot

    Inside the old gas pipeline where Ukraine is fighting an underground war

    November 18, 2026

    Kemi and Andy put another nail in the coffin of his consensus politics | John Crace

    November 18, 2026

    A rare disease reveals a hidden flaw in the immune system

    November 18, 2026
    Facebook X (Twitter) Instagram
    Trending
    • Inside the old gas pipeline where Ukraine is fighting an underground war
    • Kemi and Andy put another nail in the coffin of his consensus politics | John Crace
    • A rare disease reveals a hidden flaw in the immune system
    • Bhutan wins $1.3m Earthshot Prize for maintaining 70% forest cover
    • Maldives in talks with Indian firm over wave energy technology
    • America’s ranchers are the last holdout against industrial animal agriculture. Is their future at risk?
    • ICJ to Deliver Landmark Verdict in Myanmar Genocide Case
    • Video of red lights in the sky predates suspected drone incident at RAF Lakenheath – Full Fact
    • About
      • Our Team
      • Editorial Policy
      • Editorial Independence
      • International Support
    • Trust & Standards
      • AI Usage Policy
      • Conflict of Interest Policy
      • Corrections Policy
      • Ethics Policy
      • Fact-Checking Policy
      • Source Protection
    • Get Involved
      • Guide for Sources
      • Support Independent Journalism
    • Legal
      • Cookie Policy
      • Privacy Policy
      • Terms of Use
    Facebook X (Twitter) Instagram
    NCIJ Network NCIJ Network
    Wednesday, November 18
    • Home
    • World
    • Ai
    • Business
    • Politics
    • Health
    • Crypto
    • Science
    • Technology
    • Cybersecurity
    • Defense & Security
    • Economy
    • Energy
    • Europe
    • More
      • Fact Check
      • Investigations
      • Opinion & Analysis
      • Environment
    NCIJ Network NCIJ Network
    Home»Artificial Intelligence

    Liquid AI Releases d1.2 and d1.2-ultra: Omnimodal Decision Models That Top the AA Decision Index

    NCIJ NETWNCIJ NETWORKBy NCIJ NETWNCIJ NETWORKNovember 17, 2026 Artificial Intelligence No Comments4 Mins Read
    Share
    Facebook Twitter LinkedIn Pinterest Email

    Liquid AI just released d1.2 and d1.2-ultra. A decision model does not write text. It returns calibrated probabilities over a fixed set of answers. The Liquid AI d1.2 decision model family now accepts text, images and audio. It also adds 5 new answer types. d1.2-ultra scores 75 on the Artificial Analysis Decision Index v2.1, the top score among decision models on the index. You can check the deployment options and specs here through our GPULeaderboard.com chrome extension.

    TL;DR

    • Size: Parameters not disclosed. Up to 65,536 input tokens per question, 128 questions per call.
    • Runs on: Hosted Liquid API only. No weights yet; Liquid AI says it will open them “at some point”.
    • Performance: d1.2-ultra 75 and d1.2 69 on the AA Decision Index v2.1.
    • Best: 0.8-point calibration error for d1.2-ultra across 13 benchmarks (Liquid AI test).

    What is Liquid AI d1.2?

    d1.2 is a hosted decision model from Liquid AI that answers typed questions about text, images or audio in 1 call. It returns a probability per option and generates zero output tokens, so only input is billed.

    d1.2-ultra targets harder decisions. d1.2 is faster and costs less than half as much. The model IDs are d1.2 (default) and d1.2-ultra. Earlier d1, d1.1 and d1-ultra are now deprecated and do not support audio or the new tasks.

    What are the 5 new tasks in d1.2?

    d1.2 adds 5 typed answers on the same endpoint and request format:

    • Read a value (text, number, date): returns the value plus the characters it came from. It answers “not found” instead of guessing.
    • Find every span (spans): returns each matching passage, such as names or emails, with its position.
    • Locate objects (box, boxes): returns 1 rectangle, or 1 per instance with a count.
    • Estimate a number (number with range): returns a best guess and a range, for example 12 within 10 to 14.
    • Pick every option that applies (choice with multiple): returns each matching label with its own probability.

    How does audio input work?

    Audio can be the subject of a question, such as which language is spoken or whether a siren is audible. The question can also be spoken. The API takes 16-bit PCM WAV and counts 10 tokens per second of audio. On held-out ESC-50 and FLEURS (43 languages) recordings, both models answer 95% to 97% of questions correctly.

    How does d1.2 score on the Artificial Analysis Decision Index?

    The index combines 6 evaluations. AA-Paperclip carries 30% of the weight. EvalSafe-O*NET-AA, AA-Omniscience-Decision v1.1, HLE-MultiChoice v1.1 and AA-Symmetry v1.1 carry 15% each. When2Call carries 10%. Its Decision Score penalizes confident wrong answers.

    On the AA chart shared by Liquid AI, d1.2-ultra scores 75. Drex Max v1.1, SubQ Snap 1.2 and GLiDE v1.2 follow at 73. d1.2 scores 69. That is 2 points below the deprecated d1-ultra at 71. Liquid AI argues no model scores higher than d1.2 for less money.

    How well calibrated are the probabilities?

    A calibrated model that says 80% is right about 80% of the time, so teams can route low-confidence cases to humans. Across 13 benchmarks and 22,505 questions, Liquid AI reports these calibration errors:

    • d1.2-ultra: 0.8 points
    • d1.2: 1.2
    • Jev 1.14: 2.7
    • Perplexity Decider v1.3: 3.7
    • GPT-6 Luna (Decisions API): 7.2

    Note: These figures come from Liquid AI’s own test.

    How accurate are the new tasks?

    Liquid AI used held-out splits of public datasets; no public benchmark exists yet.

    Task (dataset) d1.2-ultra d1.2
    Read a value from receipts (CORD, SROIE) 98% 96%
    Find personal data, precision / recall (Nemotron PII) 94% / 93% 92% / 88%
    Locate an object (RefCOCO, RefCOCO+, RefCOCOg) 94% 94%
    Count objects within 1 (COCO) 93% 92%
    Pick every object present (COCO) 87% 90%

    d1.2 beats d1.2-ultra on multi-label picks, 90% to 87%.

    d1.2 vs other decision models

    Metrics d1.2-ultra d1.2 Drex Max v1.1 (Nace.AI) Decider v1.3 (Perplexity) Jev 1.14 (TypeSafe)
    AA Decision Index v2.1 75 69 73 70 66
    Calibration error (Liquid test) 0.8 1.2 Not reported 3.7 2.7
    Input price per 1M tokens $0.042 $0.02 $0.03 $0.02 $0.042
    Output price $0 $0 $0 $0 $0
    Input limit 65,536 per question 65,536 per question 262,144 262,144 65.5K
    Inputs Text, image, audio Text, image, audio Text, image Text, image Text
    Parameters Not disclosed Not disclosed Not disclosed 27B base (Qwen3.8-27B) Not disclosed
    Weights Planned, no date Planned, no date v1.1 not released Apache 2.0 Proprietary

    d1.2 d1.2ultra Decision Index Liquid models OmniModal Releases top
    NCIJ NETWNCIJ NETWORK
    • Website

    Keep Reading

    New LoRA Adapter Lifts Qwen3.8 Coding Success From 39% to 65%

    In Western Pennsylvania, You Might Live Next to—or on Top of—an Unplugged Well

    Anthropic Adds Code Mode to Claude Code: Claude Now Chains Tool Calls in JavaScript

    SID AI Releases SID-2 Agentic Search Model With 0.720 NDCG Swarm Mode

    Clearing Up the Confusion on AI Data Retention: Two Standard Designs Explained

    Perplexity Computer Adds Multiagent: Up to 16 AI Agents That Cross-Check Work

    Add A Comment
    Leave A Reply Cancel Reply

    Editors Picks

    Inside the old gas pipeline where Ukraine is fighting an underground war

    November 18, 2026

    Kemi and Andy put another nail in the coffin of his consensus politics | John Crace

    November 18, 2026

    A rare disease reveals a hidden flaw in the immune system

    November 18, 2026

    Bhutan wins $1.3m Earthshot Prize for maintaining 70% forest cover

    November 18, 2026
    Latest Posts

    After one week wearing the Pixel Watch 5, here’s what I want most

    August 21, 2026

    Targeted by an Islamophobic Provocateur, a City’s Muslims Are on Edge

    August 21, 2026

    Back on track: can Burnham fix the north’s woeful rail services? | Rail industry

    August 21, 2026

    Subscribe to News

    Get the latest sports news from NewsSite about world, sports and politics.

    NCIJ Network is an independent digital news platform delivering trusted investigative journalism, European and global news, in-depth analysis, and fact-based reporting with accuracy, transparency, and integrity.

    Facebook X (Twitter) Instagram Pinterest YouTube

    Inside the old gas pipeline where Ukraine is fighting an underground war

    November 18, 2026

    Kemi and Andy put another nail in the coffin of his consensus politics | John Crace

    November 18, 2026

    A rare disease reveals a hidden flaw in the immune system

    November 18, 2026

    Subscribe to Updates

    Get the latest creative news from FooBar about art, design and business.

    Type above and press Enter to search. Press Esc to cancel.