Sakana opens enterprise API platform early access
- Source
- sakana.ai
- Author
- Sakana AI
- Date

Context
Sakana AI is adding the account controls companies expect before they let teams call its models. The company's announcement opens early access to Sakana API Platform, which serves Sakana's LLMs through an API together with management features. An organization can create workspaces under its account, configure each one for a team or environment, and manage members, API keys and usage from one admin console. The platform also offers role-based access control, audit logs that download as CSV, budget limits, invoice payment, and a master services agreement and data processing agreement (MSA and DPA). Its data handling terms state that customer inputs are not used for training.
The models themselves are not new; the way they are sold is. Until now Sakana offered its Fugu and Namazu models as standalone APIs without management features (earlier coverage: the Fugu API launch and Namazu's release). Sakana describes Fugu as an orchestration foundation model that assembles a combination of multiple frontier models suited to a customer's workload, building on Sakana research accepted at ICLR 2026 (Trinity and Conductor). Namazu is an specialized in Japanese. Fugu and Fugu Ultra are available on the platform first, with Namazu and Fugu Max to follow. The API is OpenAI-compatible, so an application that already calls an LLM API in that format can switch to Sakana by changing the endpoint rather than rewriting its calls.
Access is gated and the commercial terms are not public. Sakana says it will admit applicants as it becomes ready to serve them, contacting each one individually after reviewing the application, and will share pricing and terms of use at that point. It also says more granular permission controls at the organization level are planned, and the admin console shown in the announcement uses demo data that may change.
- LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
Checking sign-in…
Loading comments…





