Baseten has released joey-1, a 2B parameter decision model developed by its in-house research lab, Base Labs. The new model returns a label and calibrated probability instead of generating free-form text, targeting high-volume routing, classification, and guardrail checks. Baseten has priced joey-1 at $0.01 per million input tokens and published its open weights on Hugging Face under the MIT license.
Built by post-training Qwen3.5-2B on 6.21 million decision examples using 16 H100 GPUs, joey-1 supports choices, yes/no queries, and ordered scores in a single forward pass. Baseten reports a P50 end-to-end latency of 20 milliseconds for an average 2,000-token request. On the Jev Decision Index 0.4 public split, the model achieves a score of 63.30, ranking it as the top model under 4B parameters.
This release matters because it provides developers with an ultra-fast, low-cost routing and classification tool that significantly undercuts larger models while maintaining high calibration and performance standards.