Baseten has released joey-1, a 2-billion parameter open-weight AI decision model that rapidly produces yes-or-no or multiple-choice answers along with probability scores. Priced at just $0.01 per million tokens, joey-1 provides responses in about 20 milliseconds, positioning it as an efficient alternative for companies needing high-speed, low-cost decision automation.
- 2B parameter open decision model with fast, probability-based answers
- Costs $0.01 per million tokens, answering in 20ms
- Targets use cases like support ticket routing and spam detection
What happened
Baseten has unveiled joey-1, a new AI decision model featuring 2 billion parameters and open-access weights. Unlike conventional large language models that generate textual responses, joey-1 returns direct answers with attached probability scores, enabling fast and efficient decision-making outputs.
The model is priced affordably at $0.01 per million tokens and can process typical requests in about 20 milliseconds, making it significantly faster and cheaper than Baseten’s recently released larger models mammoth-1 and elephant-1. This release expands Baseten’s portfolio of open-weight AI models focused on enterprise decision tasks.
Why it matters
Joey-1 is designed for high-volume, low-latency applications where speed and cost-efficiency are crucial, such as routing millions of support tickets monthly or identifying spam content. The model’s decision-based approach avoids text generation complexity, helping companies streamline workflows and reduce AI processing expenses.
Baseten’s claim of joey-1 being the strongest decision model under 3 billion parameters—supported by its leading rating on the Jev Decision Index for models smaller than 4 billion parameters—positions it as a competitive tool in the AI decision-making market. Additionally, the model’s open weights under an MIT license encourage broader adoption and further development.
What to watch next
Baseten’s continued expansion of decision model offerings with joey-1 signals potential growth in adoption among enterprises demanding efficient AI-driven operational solutions. Observers should monitor Baseten’s performance and user uptake, especially in sectors like customer support and content moderation where automated decision accuracy and speed are vital.
Further benchmarking and validation of joey-1’s decision accuracy and scalability will be important to corroborate Baseten’s positioning. Additionally, developments in competing decision models from players like Microsoft, which also uses Alibaba’s Qwen architecture but at a higher price point, may influence market dynamics and pricing strategies.