generated: 2026-07-21
method: searched
source: https://tinker-docs.thinkingmachines.ai/tinker/data-model/
name: Thinking Machines (Tinker) data model
description: >-
  The Tinker resource hierarchy and access-control boundaries, as documented in
  the Data Model & Permissions guide. The same model governs both the Tinker
  Console and the SDK.
entities:
- name: Organization
  description: >-
    Top-level Tinker account. Billing, user management, and org-wide settings
    live here. Organizations are fully isolated from each other; a user may
    belong to many.
  relationships:
  - has_many: User
  - has_many: Team
  - has_many: Project
- name: User
  description: An individual member of an organization, identified by email.
  relationships:
  - belongs_to: Organization
- name: Team
  description: >-
    A group of users within an org, used to grant project access to multiple
    users at once.
  relationships:
  - belongs_to: Organization
  - has_many: User
- name: Project
  description: >-
    The data-isolation boundary. Sessions, training runs, and checkpoints belong
    to a project; access is governed by project membership.
  relationships:
  - belongs_to: Organization
  - has_many: Session
- name: Session
  description: A connection between the SDK and Tinker; one per ServiceClient.
  relationships:
  - belongs_to: Project
  - has_many: TrainingRun
  - has_many: SamplingClient
- name: TrainingRun
  description: One per TrainingClient; one or more per session.
  relationships:
  - belongs_to: Session
  - has_many: Checkpoint
- name: Checkpoint
  description: A saved model/weights state; one or more per training run.
  relationships:
  - belongs_to: TrainingRun
- name: SamplingClient
  description: An inference/sampling client; one or more per session.
  relationships:
  - belongs_to: Session
permissions:
  model: project-scoped RBAC
  note: >-
    If the Console shows a checkpoint you can load it via the SDK; if a role
    denies an action in one, it denies it in the other.
docs: https://tinker-docs.thinkingmachines.ai/tinker/data-model/