Both models are free to use for one week, with the stated free-access period ending October 27.
Mammoth-1 is a 27B-parameter model and is called as baseten/mammoth-1; Elephant-1 is a 9B-parameter model called as baseten/elephant-1. Each has a 16,000-token context window. Vercel describes them as models for text classification, request routing, and scoring content against criteria set by the application.
The models return choices, probabilities, and scores. A single request can ask multiple questions about the same input, including Boolean, choice, and score questions. Applications choose the probability threshold for automation or review. Decision calls appear in AI Gateway logs and count toward its budgets alongside other model requests.
The source cites external decision-index rankings for both models, but does not provide evaluation methodology, benchmark results, pricing after the free period, rate limits, regional availability, or performance comparisons for a specific production workload.