By No Code MBA · Facts checked
- Provider
- Base Labs
- Announcement date
- Status at last check
- Generally available
- Capabilities
- Text
What changed
mammoth-1 (27B, post-trained from Qwen3.5-27B) and elephant-1 (9B, post-trained from Qwen3.5-9B) are decision models from Base Labs, Baseten's research team. Instead of generating text, they return a probability distribution over answers for choice, score and yes/no questions. They were fine-tuned on 6.21 million decision examples, including examples of when to abstain. Baseten reports mammoth-1 ranks first on Jev Decision Index 0.4 with 71.24, and scores 65.9 on the Artificial Analysis Decision Index versus 65.2 for elephant-1.
Access and availability
Both models are available through Baseten Model APIs (for example decision-models/baseten-mammoth-27b-latest) or as dedicated deployments, with zero data retention by default. Weights are on Hugging Face under the baselabs organization; Baseten's model page lists the MIT license for mammoth-1.
Official access documentationWhat you can build
Compare them on a small set of your own routing, classification, or rubric-scoring cases, and set a review threshold for uncertain outputs before automating decisions.
Limitations
They only answer fixed-format questions and do not produce free-form text. Baseten advises checking confidence calibration and escalation thresholds on your own cases, and context length was not stated.