SB Intuitions, based in Minato Ward, Tokyo, began offering a free beta version of its Sarashina service on Oct. 13, 2026, giving users a chance to try the Japan-developed generative AI Sarashina. It offers two features: web-based chat and an API that can call six AI models. Anyone can use it for free without applying in advance by logging in with a Google account.
Six tools beyond the chat
The first feature worth trying is the web chat, which lets users interact with AI and try out translation directly in a browser. But the service’s real highlight may be its API, which works behind the scenes. An API is an interface that lets users call AI functions from their own systems or apps.
The service offers six models. Sarashina3 mini handles text generation and conversation; Sarashina3 nano performs simple tasks quickly and at scale; and Sarashina3 embedding is designed for document search and RAG. RAG, or retrieval-augmented generation, is a system in which AI searches materials such as internal documents and uses what it finds to formulate an answer. The lineup also includes rerank, which improves the relevance of search results; guard, which checks the safety of input text; and translate, which specializes in Japanese-to-English and English-to-Japanese translation.
Rather than entrusting everything to one “smart AI,” the lineup divides tasks among models that write, search, reorder, monitor and translate. To use a cooking analogy, it’s less like being handed one all-purpose knife and more like having access to a kitchen stocked with a full set of tools for different jobs.
Free API could change how companies try out AI
If the service offered only chat, individuals might simply see what kinds of answers it gives and leave it at that. With the API also available, companies can integrate it into their own systems and check whether it suits their intended uses. For those considering applications such as internal help desks, document search or multilingual support, the barrier to evaluating it should be lower.
The models’ division of tasks also offers useful insights for companies considering adoption. For example, a system for searching internal documents could use embedding to make documents easier to find, rerank to refine the order of search results, and mini to generate answers. Adding guard could also build in checks on the safety of input. This is just one possible use suggested by the lineup, but choosing models according to the task at hand looks set to be key to using AI going forward.
Because the Sarashina series was developed as a large language model (LLM) with strong Japanese-language capabilities, the beta gives developers and companies a chance to try it and judge its performance on Japanese-language tasks. The company also describes a tokenizer designed for Japanese—a system that divides text into units AI can process—and says it can represent the same text using half as many tokens as U.S. models. Fewer tokens can reduce processing demands and costs, making this an important consideration for people using AI in Japanese.
Conditions to check before trying it
Although the service is free, it comes with conditions. There are limits on the number of requests users can make within a given period, so it is not suited to sending large volumes of data all at once. It is intended for small-scale trials to start.
The end date for the beta has not been announced. The company also says it may use user data to improve quality. Before trying the service for work, it would be prudent to check company policies and decide whether confidential or personal information may be entered.
SB Intuitions is a SoftBank subsidiary that conducts research and development in generative AI, focusing on the Sarashina series. The company says it will use usage data and feedback gathered during the beta to inform future product development and feature additions. By making a Japan-developed AI available for anyone to try, this launch also opens the door to a public experiment in which users’ reactions can directly inform future improvements.
Comments