OpenCOReTechnologies/CORe-Pico-4

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 16, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

CORe Pico 4 is a 1.72 billion parameter transformer decoder model developed by CORe Technologies, featuring grouped-query attention and a 40,960-token context length. Optimized for efficient execution on laptops with GGUF quants, it excels at multi-turn conversations, conditional reasoning (QuickThink), and structured tool calling. Its primary use cases include direct Q&A, agentic tasks, and identity-aware responses, making it suitable for applications requiring predictable logic and tool integration.

Loading preview...

CORe Pico 4 Overview

CORe Pico 4 is a medium-sized AI model from CORe Technologies, featuring 1.72 billion parameters and a substantial 40,960-token context length. Designed for efficient operation, it can run on laptops using GGUF quants. This model is built on a transformer decoder architecture with grouped-query attention and utilizes a 151,936-token BPE tokenizer with an embedded chat template.

Key Capabilities

  • Conditional Reasoning (QuickThink): The model can reason step-by-step when it deems it necessary, using a placeholder otherwise to save tokens. This feature can be toggled.
  • Tool Calling: Capable of emitting parseable <tool_call> JSON blocks for integration with external tools.
  • Multi-turn Conversations & Q&A: Handles ongoing dialogues and provides direct, consistent answers to factual questions.
  • Identity Awareness: Accurately responds to questions about its own identity and creator.
  • Optimized for Agentic Performance: Recommended settings (low temperature, Min P sampling) enhance logical predictability for agentic and coding tasks.

Good For

  • Agentic Applications: Its tool-calling capabilities and predictable logic make it suitable for building AI agents.
  • Coding Tasks: With appropriate settings, it can be configured for precise code-related interactions.
  • Conversational AI: Effective for chatbots requiring multi-turn interactions and direct answers.
  • Resource-Constrained Environments: Designed to run efficiently on consumer hardware like laptops, especially with GGUF quantization.
  • Applications requiring structured outputs: Its ability to generate JSON blocks for tool calls is beneficial for automated workflows.

Finetunes
2 models