Baseten unveils joey-1, a 2b decision model priced at $0.01 per million tokens

Baseten, a company specializing in AI infrastructure, has launched Joey-1, a 2B parameter decision model developed by its in-house research lab, Base Labs. The model is designed for high-volume routing, classification, and guardrail checks, returning labels and calibrated probabilities instead of generating text. It is priced at $0.01 per million input tokens, with open weights available on Hugging Face under the MIT license. Joey-1 is part of a series of models, including the previously released mammoth-1 (27B) and elephant-1 (9B), all of which share training data and the /v1/systemone API. Base Labs fine-tuned Qwen3.5-2B, keeping the vision encoder frozen and using block-bidirectional masks for full-attention layers. The model achieves a P50 end-to-end latency of 20 ms for 2,000-token requests, 2.5x faster than elephant-1. On the Jev Decision Index 0.4 public split, Joey-1 scores 63.30, ranking #37 of 230 models and leading under 4B parameters. It outperforms Cloudflare’s Clef and Jev 1.14 in four of five areas, though it lags behind mammoth-1 in overall performance.