Description
GLiDE, developed by Fastino, is a decision model that processes input data in the form of text, JSON, or images to provide probability-based outcomes for decision-making tasks. This model is particularly useful for classification, routing, and scoring applications. Users can access GLiDE through Fastino's hosted API, which requires an API key for authentication. The model supports various decision types, including choice, noul (yes/no), and score, with the ability to handle up to 255 options per question.
Read More
GLiDE is designed to work efficiently with long contexts, supporting up to 40,000 tokens per question prompt. It operates using a Python API and an HTTP server that communicates in the System One format. The model's hardware requirements include a GPU with at least 80 GB of memory, and it has been tested on NVIDIA RTX PRO 6000, B200, and H100 GPUs.
The model's weights occupy 52.2 GB on disk, with a peak GPU memory usage of 60.2 GB for a 39k-token question. GLiDE is optimized for performance, with features like flash-linear-attention for faster processing. It is deterministic, ensuring consistent results across similar requests.
GLiDE's limitations include a maximum of 255 options per question and a 40,000-token limit per question prompt. It does not generate text or explanations, focusing solely on decision-making based on provided options. The model is licensed under the Fastino Community License, with contributions from Qwen and other third-party components.
- CategoryAI Models & LLMs
- DeploymentCloud
- PlatformsWeb
GLiDE Highlights
Decision model for classification
Supports text, JSON, and images
API access via Fastino
Handles up to 255 options per question
40,000 tokens per question prompt
Python API and HTTP server
Deterministic results
Requires GPU with 80 GB memory
Getting Started with GLiDE
Access model page: Visit the GLiDE page on Hugging Face
Authenticate: Obtain an API key from Fastino
Configure: Set up the API with necessary parameters
Send prompts: Use the API to send decision requests
GLiDE's Use Cases
- Classification Tasks
- Routing Decisions
- Scoring Applications
- Multimodal Processing
- Internal Decision Support
