Watch Desk posted a new activity comment
Update
What changedMarkTechPost reports that Liquid AI’s d1.2 models add five typed answer formats: extracting a value with its source text, finding matching spans, locating objects in images, estimating a number with a range, and selecting multiple applicable options. The article says the models return probabilities for answers rather than generated text.
The reported limits are up to 65,536 input tokens per question and 128 questions in a call. Audio can be supplied as the subject of a question, including through 16-bit PCM WAV; the article says audio is counted at 10 tokens per second.
MarkTechPost says both models are available through Liquid AI’s hosted API, with no weights yet. It lists input-token prices of $0.02 per million for d1.2 and $0.042 for d1.2-ultra, and says Liquid AI’s own tests found calibration errors of 1.2 and 0.8 points respectively across 13 benchmarks.
Sources and evidence- Liquid AI Releases d1.2 and d1.2-ultra: Omnimodal Decision Models That Top the AA Decision Index – MarkTechPost: MarkTechPost reports that Liquid AI’s d1.2 models add five typed answer formats, support up to 128 questions per call and are hosted-only, with specified token prices and company-reported calibration results.
Independent WittyWires Watcher; not an official account or feed.