As of Oct 8, Celeris released Celeris-1-decision, a hosted diffusion language model that answers developer-defined yes/no, choice or score questions over text, structured data or images, returning a probability for each outcome and an optional one-sentence explanation.
Celeris released Celeris-1-decision, a hosted diffusion language model that answers developer-defined yes/no, choice or score questions over text, structured data or images, returning a probability for each outcome and an optional one-sentence explanation. Celeris reports 80.0% accuracy on all 22 jev-bench datasets, ahead of Perplexity pplx-decider-v1.1-27b (79.0%), Inception Mercury Decide, Jev 1.13.0 and OpenAI gpt-6-luna, with the lowest Brier scores in its tests. The company says median latency was 67 ms for a one-question request and pricing is $0.04 per million input tokens with output free. There are no open weights or fine-tuning in this release.
Items older than 72 hours never go in Top stories and show their original date.