The top ranked decision models in the world: mammoth-1 and elephant-1

Published October 21, 2026, the announcement introduces mammoth-1, a 27B-parameter model, and elephant-1, a 9B-parameter model, as open-source decision models post-trained from Qwen3.5 models by Base Labs. Unlike generative LLMs, these models produce probability distributions across predefined answers for choice, scoring, and yes-or-no tasks, enabling faster, lower-cost classification and routing workflows with configurable confidence thresholds and human escalation for uncertain cases. Mammoth-1 is presented as the top-ranked model on the Jev Decision Index 0.4 and as better suited to broad general-knowledge tasks, while elephant-1 is positioned as a leading smaller model for rapid reasoning. Both were fine-tuned on 6.21 million examples spanning reasoning, language, automation, retrieval, classification, safety, abstention, and subjective judgment, and are intended for applications such as ticket routing, phishing detection, prompt-safety checks, and routing queries among LLMs. The release recommends evaluating models based on accuracy, confidence calibration, latency, and cost, and makes both available through Baseten APIs for a one-week free trial and on Hugging Face.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.