Description
Fastino Labs operates as an applied research lab dedicated to building open-weight language and decision models, as well as tools tailored for production AI environments. The company focuses on advancing agentic AI capabilities through its innovative model families, notably GLiNER for extraction and classification, and GLiDE for complex decision-making tasks.
Read More
Their product offerings are structured around three core pillars: Models, Training API, and an autonomous training agent. Users can explore a diverse catalog of customized language, classification, extraction, and decision models developed by Fastino Labs. These models are available for download as weights or can be accessed for inference via their hosted API, providing flexibility for various deployment scenarios. The GLiDE decision models, ranging from 0.4B to 27B parameters, are highlighted for their performance, with GLiDE ranked #1 on Hugging Face's Decision Index.
The Training API allows users to bring their own data and SFT/RL loops to train open models on Fastino's managed infrastructure, aiming for cost efficiency compared to alternative solutions. For enterprise clients, an agentic post-training solution is available, offering advanced capabilities. The platform supports both encoder (GLiNER) and decoder (LLM) model training, with options for managed, interactive, and autonomous training workflows. The Auto Train feature delegates data preparation, training, evaluation, and model selection to Fastino's autonomous agent, streamlining the model development process.
Fastino Labs emphasizes open-weight models, fostering community engagement and widespread adoption, evidenced by over 55 million model downloads on Hugging Face and 6.7K+ GitHub stars. Their models are designed for diverse applications, including extraction, classification, decisions, and reasoning. The platform provides comprehensive documentation and quickstart guides to help users integrate their models and APIs, from setting up API keys to running inference calls and creating training jobs. Fastino Inc. is committed to data security, privacy, and responsible AI use, with ongoing efforts towards SOC 2 Type I/II and ISO 27001 certifications.
- CategoryAI Models & LLMs
- Pricing ModelContact sales
- DeploymentCloud
- PlatformsWeb
Fastino Labs's Core Features
Open-Weight Language Models
GLiNER Extraction and Classification Models
GLiDE Decision Models (0.4B to 27B parameters)
Hosted API for Model Inference
Configurable Training API
Managed Encoder Training Workflow
Interactive Decoder Training Workflow
Autonomous Training Agent (Auto Train)
Customized Model Development
Support for SFT/RL Loops
Infrastructure for Production AI
Multilingual Extraction Capabilities
Structured Choices and Confidence Scores for Decisions
API Key Management
Comprehensive Documentation
How to use Fastino Labs?
Create Account: Sign up for a Fastino account to get started.
Obtain API Key: Retrieve your API key from the settings section.
Run Inference: Use cURL, Python, or JavaScript to make your first API call for model inference.
Define Schema (for GLiNER): For extraction tasks, define a schema with entities, classifications, structures, or relations.
Create Training Job: Initiate a decoder training job by sending a dataset and configuration to the training API.
Monitor Training: Track the lifecycle of your training jobs, from requested to complete or deployed.
Explore Demos: Utilize the demo library to see practical applications and examples.
Integrate Models: Incorporate Fastino's models into your applications for various AI workflows.
Fastino Labs's Use Cases
- Custom Model Development
- Structured Data Extraction
- Text Classification
- Complex Decision Making
- Automated Training Workflows
- Robotics Navigation
- Live Content Analysis
- Enterprise AI Solutions
- Financial and Healthcare AI
- API-Driven Inference
