Meet bison-1 and groundhog-1, frontier American decision models
Blog post from Baseten
Base Labs has released bison-1 and groundhog-1, decision models built on Google’s Gemma 4, positioning them as American-backbone alternatives to its earlier Qwen-based models. Both accept text and image input, provide 128,000-token context windows, and are available through Baseten Model APIs with open weights on Hugging Face; their listed prices are $0.10 and $0.035 per million tokens, respectively. On Artificial Analysis Decision Index v2.1, bison-1 ranks first among 19 models with a score of 76 and has the index’s lowest Brier score, while the company expects groundhog-1 to perform comparably to mammoth-1. Base Labs says the new models offer improved quality with similar speed to their predecessors, and that its model family spans a range of accuracy, latency, and cost trade-offs. The company attributes bison-1’s calibration in part to training with soft labels and says the longer context windows help teams provide more complete documents or conversation histories without first summarizing or selecting excerpts.
No tracked trend matches for this post yet.
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.