Description
Clef is a sophisticated multimodal AI model developed to convert various input formats such as text, JSON, images, and video into actionable decisions. With a backbone based on Qwen/Qwen3.8-27B, Clef utilizes a vision encoder and a joint schema head to process and score options for questions. This model is particularly notable for its ability to handle diverse input types and provide probabilities for each allowed option without generating free-form text or parsing outputs.
Read More
The Clef API is compatible with Jev and SystemOne, facilitating seamless integration into existing systems. It is post-trained from Qwen/Qwen3.8-27B and offers a smaller, faster variant known as Clef-Flash. The model's architecture includes a transformer head that routes evidence from the state to each question, scoring all options jointly.
Clef supports image and video inputs, requiring additional libraries such as pillow for processing. It can be served with SGLang, with specific launch commands available for different hardware setups. The model has been tested extensively across various benchmarks, demonstrating high accuracy and efficiency in decision-making tasks.
Released under the Apache-2.0 license, Clef is accessible for download and integration, with detailed configuration files available for setup. Its robust performance in benchmarks like BFCL, ToolRet, and API-Bank highlights its capability in exact accuracy and decision-making speed.
- CategoryAI Models & LLMs
- Pricing ModelOpen Source
- DeploymentCloud
- PlatformsWeb
Clef Highlights
27B Multimodal Model
API Compatibility with Jev and SystemOne
Post-trained from Qwen/Qwen3.8-27B
Vision Encoder
Joint Schema Head
Image and Video Input Support
Apache-2.0 License
Clef-Flash Variant
Getting Started with Clef
Access model: Navigate to the Hugging Face page
Authenticate: Ensure API access credentials
Set up environment: Install necessary libraries like torch and pillow
Integrate via API: Use Jev/SystemOne API for requests
Optimise: Adjust configurations for performance
Clef's Use Cases
- Decision Making
- Multimodal Processing
- API Integration
- Benchmark Testing
- Image and Video Analysis
