Nimble
AI model · Safety / moderation · Text → Tables / structured data by Ollama
High-speed 9B parameter decision model for classification and scoring.
Nimble is an open-source 9B parameter decision model developed by Bespoke Labs and available through Ollama. It is specifically designed for high-speed, typed decisions such as classification, scoring, and yes/no questions. By running locally, Nimble eliminates network latency, achieving decision speeds as fast as 91ms for real-time applications like game logic or ticket triage. The model provides structured answers including confidence scores and probabilities, making it suitable for automated moderation and routing workflows.
- Price
- From 0 USD
- Languages
- English
- Parameter Count
- 9B
- Developer
- Bespoke Labs
- Input Modality
- Text/JSON State
- Output Modality
- Structured Decisions
- Average Latency
- 91ms (on M5 Max)
Overview
Nimble is an open-source 9B parameter decision model developed by Bespoke Labs and available through Ollama. It is specifically designed for high-speed, typed decisions such as classification, scoring, and yes/no questions. By running locally, Nimble eliminates network latency, achieving decision speeds as fast as 91ms for real-time applications like game logic or ticket triage. The model provides structured answers including confidence scores and probabilities, making it suitable for automated moderation and routing workflows.
Decision Capabilities
- Rapid Decision Making: The model is designed for fast, typed decisions such as ticket triage and content moderation.
- Low Latency: Nimble averaged 91ms per decision when running locally on an M5 Max processor.
- Probabilistic Outputs: The model provides probabilities and confidence scores for every answer it generates.
- Flexible Decision Types: Supports multiple decision types including choice, yes/no (noul), and scoring.
Developer Experience
- Local Execution: Requests run locally without traveling over a network, ensuring data privacy and speed.
- Typed API Support: Integrates with TypeSafe's Jev API for structured, typed decision logic.
- CLI Integration: Users can pipe text or JSON state directly into the model via the CLI.
Pricing
- Local Usage
- Free
Specifications
- Parameter Count
- 9B
- Developer
- Bespoke Labs
- Input Modality
- Text/JSON State
- Output Modality
- Structured Decisions
- Average Latency
- 91ms (on M5 Max)
Added by AntSpace from public sources (ollama.com).