header-langage
简体中文
繁體中文
English
Tiếng Việt
한국어
日本語
ภาษาไทย
Türkçe
Scan to Download the APP
Partnered With:

Liquid AI Decision Model Upgrade: Adds 5 New Capabilities, Tops the Chart with 75 Points, Price Unchanged

Beating AI News Flash: Liquid AI releases two multimodal decision models, d1.2 and d1.2-ultra.


The new models scored 69 and 75 points respectively on the Artificial Analysis decision model leaderboard, improving by 3 and 4 points over their respective predecessors. Among them, d1.2-ultra ranks first, leading the second place by 2 points.


The context window of d1.2-ultra has also expanded from 33,000 to 66,000 tokens. The average task time has been reduced from 301 milliseconds in the previous generation to 218 milliseconds, a decrease of about 28%.


Previously, d1-ultra was mainly responsible for true/false judgments, single-choice questions, and scoring. The new version adds 5 types of tasks: extracting specific values such as amounts and dates from documents; finding all qualifying text and its location; circling specified objects in images; estimating quantities and giving a range; and selecting multiple qualifying options at once.


For example, when processing invoices, the model can now directly extract the amount payable and return the corresponding original text. If it is not found, it answers "not found." When processing images, it can also box all objects that match the description, rather than just determining whether they exist.


The two models also add audio input, allowing them to directly recognize speech, sirens, and other sounds in recordings, and they also support asking questions by voice.


Liquid AI's internal tests show that d1.2-ultra achieves 98% accuracy in extracting numerical values from receipts and 94% in image object localization.


In another comparative test covering 22,505 questions, its confidence calibration error was only 0.8 percentage points, while Perplexity Decider v1.3 was 3.7 percentage points. That is, the judgment probabilities it gives are closer to the actual accuracy rate, making it easier for systems to identify tasks that require manual review.


Pricing remains unchanged. d1.2 and d1.2-ultra charge $0.02 and $0.042 per million input tokens respectively, with no output fee, and both are now available via API.


Liquid AI says it will launch open-source versions with comparable capabilities in the future.

举报 Correction/Report
Correction/Report
Submit
Add Library
Visible to myself only
Public
Save
Choose Library
Add Library
Cancel
Finish