Baseten launches Bison-1 and Groundhog-1 decision models with 128k context
Baseten adds two open‑weight decision models built on Gemma 4, offering calibrated probabilities over 128k tokens.
Both models are based on Google's Gemma 4 architecture and return calibrated probabilities instead of text.
Groundhog-1 is a 26‑billion‑parameter mixture‑of‑experts model with about 4 billion active parameters per token, but it has not yet received an official index score.
Both models support a context window of 128 k tokens (131,072) and can accept image inputs. Their BF16 weights are available on Hugging Face, sized at roughly 62 GB for Bison-1 and 52 GB for Groundhog-1.
Pricing is $0.10 per 1 M input tokens for Bison-1 and $0.035 per 1 M tokens for Groundhog-1.
No inference GPU or quantized builds are listed, and the models are open‑weight under MIT + Apache 2.0 licenses.
What it means for you
These models give developers ready‑to‑use decision‑making services that handle long contexts and images, reducing the need to build custom pipelines.
Sources
- Baseten Releases Bison-1 and Groundhog-1, Topping Decision Model Index, MarkTechPost, 21 Nov 2026
- baselabs/groundhog-1 · Hugging Face, huggingface.co (official), 21 Nov 2026
Written automatically from the sources above, then checked by our code: every number and name must appear in them, and no sentence may copy a source. Spotted a mistake? Report it.
Which AI can you use for free today?
Every free AI model, ranked by blind human votes, with the memory and speed it needs on your own GPU.