树林AI
🌐
Built for Japanese business workflows, JJ-1 delivers fast, structured decisions for AI agents and enterprise software
Jurin AI Inc., the company building AI agents for high performing enterprises, today introduced JJ-1, a new family of decision models built for structured decision-making inside enterprise workflows.
As enterprises move from AI that generates answers to AI agents that take action, more of the work inside software comes down to decisions: Which team should receive a customer request? How urgent is it? Should it be escalated? Which workflow should run next?
Across more than 88,000 questions spanning six evaluation areas, including public English- and Japanese-language datasets and Japanese enterprise workflow data, JJ-1 achieved an aggregate accuracy of 79.2%, compared with 76.5% for Jev 1.13 and 71.35% for GPT-6 Luna Decision.
Within this evaluation, JJ-1 achieved 97.43% accuracy on the Japanese enterprise workflow benchmark, compared with 91.99% for GPT-6 Luna Decision and 91.75% for Jev 1.13. JJ-1 also delivered up to 99% lower inference cost than generative LLM approaches.
Rather than generating an answer token by token, JJ-1 scores predefined options directly and returns probabilities that software can use to determine the next action.
JJ-1 is designed around three priorities:
For developers building AI agents and SaaS products, these decisions can happen many times within a single workflow. A system may classify a request, assess urgency, choose a route, decide whether to escalate, and select the next action before any work is completed. That makes latency and inference cost increasingly important as these systems scale.
Japanese business language can be highly context-dependent. “Kekkō desu” (結構です), for example, can signal either acceptance or refusal depending on the conversation.
In recorded examples, JJ-1 correctly distinguished between acceptance and refusal uses of “kekkō desu,” returning the decision in milliseconds.
What we’re seeing from deploying AI agents in Japan is that once they start doing real work, they make decisions constantly. One user request can involve several decisions before anything actually gets done, and at scale, every extra model call adds time and cost. The future isn’t one giant model doing everything. Use the best generative models when you need deep reasoning or generation. Use a model built for decisions when it’s deciding where something goes, how urgent it is, or whether to escalate. JJ-1 is what we’ve built for that specialized decision layer for Japanese enterprises.
By leveraging the infrastructure power of ASIC and GPU technology, JJ-1 democratizes the Japanese open-source ecosystem. This demonstrates that a world-class text model can be built and trained entirely on Japan-based cloud infrastructure to benefit the broader community of the world.”
Jurin AI is publishing the full benchmark methodology, testing conditions, and cost comparisons in the JJ-1 Technical Report, available on the company’s website.
JJ-1 is available to try today at inference.jurin.ai.
Jurin AI is building the AI Agent for high-performing enterprises. Our technology automates business communications (email, phone, messaging) and the workflows behind them for a better way to do business.
Within just one year of founding, we have become one of the fastest growing startups globally. With offices in Japan, Korea, and Singapore, we are steering Asia-Pacific towards an agentic future. Trusted by large Japanese enterprises every Tokyo resident relies on and respects, Jurin AI integrates seamlessly into existing tools and systems to execute tasks across 100+ languages, enabling teams to tackle labor shortage, focus on strategic, high-value activities, and finally unlock scale to grow the business by another 10x.
媒体联络
press@jurin.ai
