System One
Product · API / developer platform · Customer support by Ollama
Fast, typed decisions for classification and scoring.
System One is a specialized API for fast, typed decision-making within Ollama. It enables developers to perform real-time classification, scoring, and yes/no questions using local or cloud-hosted models.
- Launched
- 2026-09-29
Category: API Management
- APIs and SDKs
- /v1/systemone API, TypeSafe Python SDK
- Platforms
- API, CLI, Python
- Deployment
- Local (macOS, Windows, Linux), Ollama Cloud (US and Europe)
- API Endpoint
- /v1/systemone
- Local Latency (Nimble 9B)
- 91ms (M5 Max)
- Question Types
- choice, noul, score
- 91ms average latency for Nimble 9B on M5 Max
- Supports choice, yes/no (noul), and score types
- Available via /v1/systemone endpoint
- Zero data retention for cloud requests
Overview
System One is Ollama's specialized API and framework for fast, typed decision-making, based on TypeSafe's Jev API. It allows developers to perform classification, scoring, and yes/no questions by sending a text "state" with named questions in a single request. System One is designed for low-latency tasks such as ticket triage, model routing, and content moderation. It supports local execution with models like Nimble and Tev1, as well as cloud-hosted inference using the Clef model. When running locally on an M5 Max, the Nimble 9B model can average 91ms per decision, enabling real-time processing and agentic workflows.
Decision Engine
- Typed Decisions: System One performs classification, yes/no questions, and scoring in a single request.
- State-based Queries: Developers send text as state along with a set of named questions to receive structured answers.
- Probabilistic Outputs: The system provides a probability for every option and a confidence score for answers.
- Real-time Moderation: System One is optimized for tasks like ticket triage, model routing, and content or safety moderation.
Developer Access
- Unified Endpoint: The API is accessible via the /v1/systemone endpoint for both local and cloud requests.
- TypeSafe Python SDK: A dedicated Python SDK from TypeSafe is available for making requests to System One.
- Local Execution: Local requests do not travel over a network, resulting in near-instant decisions.
- Cloud Availability: Cloud access is available for classification and scoring without requiring a local GPU.
Use cases
- Developers use System One to triage incoming support tickets into billing, technical, or other categories.
- Content moderators use decision models to score the urgency or safety of user-generated text in real time.
- Engineers route requests to different AI models based on the classification of the user's prompt.
Specifications
- APIs and SDKs
- /v1/systemone API, TypeSafe Python SDK
- Platforms
- API, CLI, Python
- Deployment
- Local (macOS, Windows, Linux), Ollama Cloud (US and Europe)
- API Endpoint
- /v1/systemone
- Local Latency (Nimble 9B)
- 91ms (M5 Max)
- Question Types
- choice, noul, score
- Cloud Model Name
- clef
- Minimum Version
- Ollama v0.35
- SDK Support
- TypeSafe Python SDK
How it works
- Install Ollama: Download or upgrade to Ollama v0.35 or later and pull a decision model like nimble.
- Define State and Questions: Send a POST request to the /v1/systemone endpoint with a state object and a set of questions.
- Receive Structured Answer: The model processes the state and returns structured answers with probabilities and confidence scores.
Pricing
- Local Usage
- Free with local models
- Cloud Usage (clef model) per 1M tokens
- $0.24
FAQ
What is System One?
System One is used for fast, typed decisions such as classification, scoring, and answering yes/no questions.
Which version of Ollama supports System One?
System One is available as of Ollama version 0.35.
Can I use System One in the cloud?
Yes, System One is available in the cloud for classification, yes/no questions, and scoring, though supported capabilities vary by model.
What model does System One use in the cloud?
The cloud-hosted model for System One is named clef.
Added by AntSpace from public sources (docs.ollama.com).