# Baseten's Mammoth-1 and Elephant-1 now available on AI Gateway for free

Source: [Vercel](https://vercel.com/changelog/basetens-mammoth-1-and-elephant-1-now-available-on-ai-gateway-for-free)  
Feed7 permalink: https://feed7.dev/p/basetens-mammoth-1-and-elephant-1-now-available-on-ai-gateway-for-free-0b4xffp  
Published: 2026-10-21T00:00:00.000Z  
Trust: Official Source (official_source)

## Why Included

Baseten’s 27B and 9B decision models can classify, route, and score text with returned probabilities, giving agent workflows explicit thresholds for automation or review.

## Source Summary

Baseten’s **Mammoth-1 is 27B** and **Elephant-1 is 9B**. Both decision models accept up to **16k tokens** and return choices, probabilities, and scores for classification, routing, or criteria-based evaluation.

## Practical Implication

Use them where an agent needs a constrained decision rather than free-form prose. One request can ask multiple question types, while the application sets probability thresholds for automatic action versus human review.

## Agent-Ready Context

Baseten’s **Mammoth-1 is 27B** and **Elephant-1 is 9B**. Both decision models accept up to **16k tokens** and return choices, probabilities, and scores for classification, routing, or criteria-based evaluation.

Use them where an agent needs a constrained decision rather than free-form prose. One request can ask multiple question types, while the application sets probability thresholds for automatic action versus human review.

Free AI Gateway access was limited through **October 27**. The ranking claims come from two different decision indexes, so their number-one placements should not be treated as a direct head-to-head comparison.

## Connected Context

Feed7 judgment across 898 accumulated Signals:

This narrows model selection from general generation to constrained, probabilistic decisions suitable for routing, classification, and review gates. The 27B and 9B options create a size choice, but separate ranking indexes do not establish which is better. Applications must define thresholds and escalation policy, and the temporary free access supports evaluation rather than a durable cost assumption.

- [Jev is the fastest-adopted model in AI Gateway history](https://feed7.dev/p/ai-gateway-jev-model-launch-0rvbmm3) — Jev provides the closest architectural precedent: both return typed probabilistic decisions that can sit around general agents as routing, guardrail, or approval components.
- [When Does Bigger Help? A Controlled Study of LLM Scale for Ontology Learning](https://feed7.dev/p/2608-31118v1-1abyqwd) — The controlled finding that model size is a weak selector cautions against choosing Mammoth-1 over Elephant-1 from 27B versus 9B alone; task-specific decision quality still needs testing.
- [Kimi K3.1 Pro and Kimi K3.1 Pro Highspeed are available now in AI Gateway](https://feed7.dev/p/kimi-k3-1-pro-and-kimi-k3-1-pro-highspeed-are-available-now-in-ai-gatewa-1gne8bl) — Kimi represents a general long-context agent model, whereas Mammoth and Elephant are specialized decision layers that could route or evaluate such models rather than replace their free-form work.

## Context Map

- Layer: model
- Domains: data
- Topics: model-selection

## Uncertainty

- Free AI Gateway access was limited through **October 27**. The ranking claims come from two different decision indexes, so their number-one placements should not be treated as a direct head-to-head comparison.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.