d1-ultra
A multimodal decision model released by Liquid AI, designed specifically for classification, routing and scoring tasks. It accepts text and images as input and outputs structured probabilistic judgments rather than generated text.
Core Facts
Timeline (last 90 days)
d1-ultra claims to output calibrated probabilities in a single forward pass, and the authors speculate that it may embed the calibration mechanism directly into the training objective or output layer design rather than relying on post-hoc correction.
According to a MarkTechPost report, d1-ultra scored 65 on the Artificial Analysis Decision Index, with an input cost of $0.042 per million tokens.
d1-ultra supports three question types: Boolean (yes/no probability), Choice (closed-set multi-class classification), and Score (ordinal regression / ordered scale scoring).
d1-ultra supports posing multiple questions about the same input within a single request and returning their respective probabilities.
The core capability emphasized officially by d1-ultra is calibrated probabilities, meaning the model's output probabilities should match actual occurrence frequencies.
d1-ultra supports a context window of 32,768 tokens.
Typical scenarios supported by d1-ultra include visual defect inspection and comparison of customer photos with damage claims.
In visual defect detection scenarios, threshold rules can be set based on calibrated probabilities: probabilities above 0.95 are automatically passed, those below 0.6 are directly rejected, and the intermediate range is routed to manual review.
d1-ultra's billing method charges only for input tokens, does not charge output token fees, and image inputs also count toward input tokens.
d1-ultra is a multimodal decision model specifically handling classification, routing, and scoring tasks; its input supports text and images, and it does not generate text but rather outputs structured probabilistic judgments.
1 more timeline events
All Facts (11)
Liquid AI has expanded its D1 decision model with visual capabilities, enabling it to accept images, text, or both, and return probabilities for yes/no judgments, multiple-choice, or scoring questions.
70%Unverifiedd1-ultra claims to output calibrated probabilities in a single forward pass, and the authors speculate that it may embed the calibration mechanism directly into the training objective or output layer design rather than relying on post-hoc correction.
50%UnverifiedAccording to a MarkTechPost report, d1-ultra scored 65 on the Artificial Analysis Decision Index, with an input cost of $0.042 per million tokens.
50%Unverifiedd1-ultra supports posing multiple questions about the same input within a single request and returning their respective probabilities.
50%UnverifiedThe core capability emphasized officially by d1-ultra is calibrated probabilities, meaning the model's output probabilities should match actual occurrence frequencies.
50%Unverifiedd1-ultra's billing method charges only for input tokens, does not charge output token fees, and image inputs also count toward input tokens.
50%Unverifiedd1-ultra supports a context window of 32,768 tokens.
50%UnverifiedTypical scenarios supported by d1-ultra include visual defect inspection and comparison of customer photos with damage claims.
50%UnverifiedIn visual defect detection scenarios, threshold rules can be set based on calibrated probabilities: probabilities above 0.95 are automatically passed, those below 0.6 are directly rejected, and the intermediate range is routed to manual review.
50%Unverifiedd1-ultra is a multimodal decision model specifically handling classification, routing, and scoring tasks; its input supports text and images, and it does not generate text but rather outputs structured probabilistic judgments.
50%Unverifiedd1-ultra supports three question types: Boolean (yes/no probability), Choice (closed-set multi-class classification), and Score (ordinal regression / ordered scale scoring).
50%