From the source
Baseten released mammoth-1 and elephant-1, two decision models trained by its Base Labs research team, which return probability scores instead of generating text.
- The company says mammoth-1 ranks #1 of 208 decision models and elephant-1 leads 9B decision models on the Jev Decision Index, while elephant-1 ties for #1 with d1-ultra on the Artificial Analysis Decision Index.
- Baseten post-trained Qwen3.5-9B and Qwen3.5-27B on 6.21 million decision-making examples drawn from public and synthetic data covering decision skills, safety and abstention, and judging and expert domains.
- Both models are open-source on
Hugging Face and free to try for a week through Baseten Model APIs, with zero data retention by default.
- Decision models take choice, score or boolean questions and are served without text generation, which Baseten says can be up to 10x faster than a reasoning model and cheaper than using a generative LLM for classification.
