kedarcv/Clair-3B

TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jun 29, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Clair-3B by Michael Mlungisi Nkomo is a 3.09 billion parameter personalized AI assistant, fine-tuned from Qwen2.5-3B-Instruct with a 4096-token context length. It features an embedded identity, allowing it to maintain consistent persona across interactions. Optimized for efficient operation on budget laptops (CPU-only, 8GB RAM), Clair-3B serves as a versatile assistant for coding, math, writing, analysis, and general questions.

Loading preview...

Overview

Clair-3B is a 3.09 billion parameter personalized AI assistant developed by Michael Mlungisi Nkomo, fine-tuned from the Qwen2.5-3B-Instruct base model. Its core differentiator is an embedded identity, ensuring consistent persona and responses across all interactions. This model is specifically engineered for efficient performance on budget hardware, including CPU-only laptops with 8GB RAM, making it accessible for a wide range of users.

Key Capabilities & Features

  • Embedded Identity: Maintains a consistent persona named "Clair," explicitly denying being other LLMs like ChatGPT or Claude.
  • Resource-Efficient: Designed to run on low-spec hardware, with quantized versions (e.g., Q4_K_M) requiring approximately 2.5 GB RAM.
  • Versatile Assistance: Functions as an AI assistant for coding, mathematical tasks, writing, data analysis, and general inquiries.
  • Optimized Training: Fine-tuned using LoRA (rank 32, alpha 64) over 20 epochs on a dataset emphasizing identity and multi-turn dialogues, achieving 100% identity recognition.
  • Flexible Deployment: Supports integration with popular frameworks like Hugging Face Transformers, Ollama, and llama.cpp (GGUF).

Ideal Use Cases

  • Personal AI Assistant: Users seeking a consistent, personalized AI companion for daily tasks.
  • Edge Device Deployment: Applications requiring an LLM to run efficiently on resource-constrained devices.
  • Educational & Development: Developers and students working with AI on budget laptops or limited hardware.
  • Interactive Applications: Scenarios where maintaining a distinct AI persona is crucial for user experience.

Finetunes
2 models