- input
- $0.10
- output
- $0.00
- Context
- 16Ktokens
- Released
- 3w ago
- Size
- 27.4Bparameters
Cost/1M tokens
Mammoth 1 is an open-source decision model from Baseten's Base Labs research team, post-trained from Qwen3.5-27B and released alongside the smaller elephant-1. Baseten's announcement calls the pair the world's top-ranked decision models, claiming mammoth-1 ranks first out of 208 decision models on the Jev Decision Index. Unlike generative LLMs, it returns a probability distribution over possible answers instead of text, and is built for decision tasks such as ticket triage, spam and phishing detection, guardrails, and routing requests between small and frontier LLMs.
- Decision model
- Open weights · 27.4B
- Accepts images
How it compares
Family tree
Models trained from one another, as their makers describe it.
Qwen3.5-27BQwen · 2026
Frequently asked questions
Running locally
Mammoth 1 needs an estimated 19.6 GB of VRAM at Q4, 33.9 GB at Q8 and 63.4 GB at FP16, based on its 27.4B parameters. All figures are estimates that include weights plus about 15% and 0.5 GB for a short context.
Yes, a 24 GB GPU fits at Q4 with an estimated 19.6 GB. A 16 GB, 12 GB or 8 GB GPU does not fit at Q4, Q8 or FP16 (estimate).
Yes. The smallest common Mac memory size that fits Q4 is an estimated 32 GB, assuming macOS lets the GPU use about 75% of unified memory, since Q4 needs an estimated 19.6 GB.
API and pricing
Input costs $0.10 per 1M tokens and output costs $0.00 per 1M tokens at the 1 paid provider tracked.
A typical request with 10,000 input tokens and 1,000 output tokens costs $0.001 at the cheapest provider, BaseTen.
The 1 endpoint with uptime data in the last day ran at 99% uptime or better, with the best uptime of 100.0% by BaseTen.
Privacy and data
BaseTen offers zero data retention (ZDR eligible or 0-day retention).
Yes, Mammoth 1 has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 16K tokens (16,384) and the maximum output is 15K tokens (14,745).
Mammoth 1 can read text and image inputs. Audio, video and PDF are not listed as input modalities.
Quality and benchmarks
Clairic by Cloudflare at $0.075 per 1M tokens, Clef Omni by Cloudflare at $0.1125 per 1M tokens and Drex v2.0 by Nace.AI at $0.045 per 1M tokens, matched on model kind, blended price and release date.
ClairicMammoth 1 vs ClairicClef OmniMammoth 1 vs Clef OmniDrex v2.0Mammoth 1 vs Drex v2.0
Versions and lineage
Mammoth 1 is fine-tuned from Qwen3.5-27B.
Licensing and access
Mammoth 1 has open weights.
Download the weights from the Hugging Face repository baselabs/mammoth-1.
Basics
Baseten made Mammoth 1, released October 21, 2026.
No. The weights are open, but the listed API price is $0.10 input and $0.00 output per 1M tokens.
Generated from Vertical Tabs data, updated November 11, 2026.
Discussion
No comments yet
Be the first to share your thoughts.