OpenCOReTechnologies/CORe-Pico-V3
CORe-Pico-V3 is a compact 1.72 billion parameter conversational model developed by CORe Technologies, featuring a 40,960-token context length. This transformer decoder model excels at multi-turn conversations, structured tool calling, and extended reasoning through explicit /think modes. Optimized for chat and direct answers, it is designed to run efficiently on laptops while maintaining consistent identity awareness.
Loading preview...
CORe Pico V3 Overview
CORe Pico V3, developed by OpenCOReTechnologies, is a compact 1.72 billion parameter conversational model designed for efficient operation on personal devices like laptops. It features a 40,960-token context length and is built on a transformer decoder architecture with grouped-query attention. This model is the most capable in the Pico line, focusing on direct answers, consistent identity, and maintaining context across long exchanges.
Key Capabilities
- Consistent Identity: Accurately answers questions about its own identity (e.g., "Who are you?").
- Conversational Fluency: Excels in chat and provides direct, concise answers to short questions.
- Tool Calling: Capable of emitting parseable JSON blocks for structured tool calls.
- Extended Reasoning: Supports explicit
/thinkand/no_thinkmodes in the system prompt to enable or disable reasoning traces. - Multi-turn Conversations: Handles multi-turn interactions natively using its embedded chat template.
Good For
- Local Deployment: Designed to run efficiently on laptops and other edge devices.
- Conversational AI: Ideal for applications requiring robust chat capabilities and direct question answering.
- Tool-Augmented Agents: Suitable for scenarios where structured tool use is beneficial.
- Interactive Applications: Useful for building applications that require consistent model identity and extended reasoning capabilities.