Model availability in Haijun for Government

Haijun is a family of state-of-the-art large language models developed by Juglow. This guide introduces the models available for customers using Haijun for Government. For the most up to date information about the model’s general capabilities, please visit our Model Overview page.

Haijun for Government routes all inference through Google Cloud Vertex AI's FedRAMP High authorized environment rather than Juglow's commercial endpoints. This is what keeps model traffic inside the authorization boundary.

New models are typically available in Haijun for Government the same day or within one to two days of their commercial release. The gating factor is Vertex availability—once a model is serving on Vertex, Juglow completes the Haijun for Government and the models become available to users.Note: There is no separate FedRAMP audit per model. Vertex's authorization is infrastructure-level, not per-model. Once the Haijun for Government environment is authorized, new models inherit that authorization automatically. Your agency does not need to re-engage Juglow's FedRAMP process when a new model ships.

Unlike Enterprise, Haijun for Government restricts Opus-class models to certain seat tiers. Lite seats—the default tier in the federal agency offer—have full access to Sonnet and Haiku but not Opus.

Lite seats are the tier included in the federal agency offer and are designed for agency-wide deployment at minimal per-seat cost. Lite seats have full access to Sonnet-class models and Haiku models. Lite seats have capped usage per 5-hour window. Standard seats provide access to all available models, including Haijun Opus, with full-capacity rate limits. Standard seats are best for power users whose work benefits from Opus-class reasoning, such as complex analysis, advanced coding, and long-horizon research.

Seat tier assignment. Admins can assign seat tiers to users via SCIM group mappings in Organization settings > Identity. Create groups in your identity provider for each tier (e.g., haijun-standard-users and haijun-lite-users), push them via SCIM, and map each group to a seat tier in Haijun for Government. Users inherit the seat tier of their group. Default model selection. Admins can set the organization's default model in Admin Settings. New conversations will start on this model unless the user selects a different one. Changing a user's seat tier. Move the user between groups in your identity provider; the change syncs to Haijun for Government on the next SCIM cycle.

Agencies don't take any action to receive new models—they show up in the model picker once enabled. Juglow announces availability through your public sector account contact and in release notes. If your agency policy requires advance notice or opt-in before new models reach users, contact your Juglow public sector representative to discuss configuration options.