Mammoth-1 and Elephant-1

Baseten's open decision models (27B and 9B) that return probabilities for choice, score, and yes/no questions instead of text.

Baseten · Open-weight/community models · model

Published in Feed

This entry is marked for review. Check the original sources before relying on its details.

Overview

Baseten released mammoth-1 (27B) and elephant-1 (9B) on October 21, 2026 as open-weight decision models that return a probability distribution over answers rather than generating text. Per Baseten's blog they were post-trained from Qwen3.5-27B and Qwen3.5-9B; both support a 16k-token context. Baseten reports mammoth-1 scoring 71.24 on the Jev Decision Index 0.4 and 65.9 on the Artificial Analysis Decision Index (elephant-1: 65.2), vendor-reported and not independently verified. They are served through Baseten Model APIs (System One endpoint) and offered free on Vercel AI Gateway through October 27.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Decision outputs: choice, score, and yes/no probabilities, multiple questions per request
  • mammoth-1 27B and elephant-1 9B, 16k-token context
  • Vendor-reported 71.24 (mammoth-1) on Jev Decision Index 0.4; 65.9 and 65.2 on Artificial Analysis Decision Index
  • Served via Baseten Model APIs System One endpoint with zero data retention by default
  • Free on Vercel AI Gateway through October 27, 2026

Use cases

  • Text classification and request routing
  • Scoring content against a rubric in one low-latency call

Pricing

Free to try for a week on Baseten Model APIs and free on Vercel AI Gateway through October 27, 2026; paid pricing and license not stated in the sources reviewed.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
21 Oct 2026
Record updated
23 Oct 2026
Record last checked
23 Oct 2026

Explore the Feed

Other entries filed under open-weight/community models; inclusion does not imply equivalent features.

Browse the Feed