From the source
Baseten's research lab Base Labs has added joey-1, a 2B decision model it calls its fastest and most affordable, eight days after releasing mammoth-1 and elephant-1.
- Decision models return a label and a calibrated confidence score in one forward pass instead of generating text, which Base Labs says cuts the cost and latency that compound across chained decisions.
- In internal testing, joey-1 is the strongest decision model with 3B or fewer parameters, and at production scale it serves an average 2,000-token request at a P50 end-to-end latency of 20 ms, about 2.5x faster than elephant-1's 49 ms.
- At $0.01 per million tokens, joey-1 costs a tenth of mammoth-1; classifying 10 million support tickets a month at roughly 500 tokens each would run about $50 on joey-1 against $500 on mammoth-1. joey-1 is post-trained from Qwen3.5-2B using the same data and recipe as the larger models, and all three share one API, so switching between them is a one-line change.
- It is available now on Baseten Model APIs with a 16k context window and zero data retention by default, with open weights on Hugging Face; dedicated deployments reach the full 262k context window and support fine-tuned weights.
