logihertz/nyra-C
Nyra-C is an 8 billion parameter instruction-following and code generation model developed by Logihertz Systems, based on an optimized Llama-3 architecture with a 32768 token context length. It is specifically engineered as an "Instruction & Execution Core" for zero-shot instruction following, agentic tool-use, and reliable output of strict formats like JSON and YAML. This model excels at backend operations, API integrations, and developer-facing tasks, including writing and debugging code across multiple programming languages.
Loading preview...
Nyra-C: Instruction & Execution Core
Nyra-C, developed by Logihertz Systems, is an 8 billion parameter model built on an optimized Llama-3 architecture. It functions as an "Instruction & Execution Core" (Tier C) within the independent Nyra Project, focusing on precision and strict adherence to instructions.
Key Capabilities
- Zero-shot Instruction Following: Designed for reliable execution of commands without prior examples.
- Code Generation: Proficient in writing, debugging, and refactoring code in Python, JavaScript, and C++.
- Strict Formatting: Ensures syntactically correct outputs for formats like JSON, YAML, and XML, minimizing conversational filler.
- Agentic Tool-Use: Optimized for integration into multi-agent frameworks, enabling intelligent routing and external tool selection.
Intended Use Cases
Nyra-C is particularly well-suited for backend operations, API integrations, and developer-centric applications where precision and structured outputs are critical. Its strengths lie in automating tasks requiring strict formatting and robust code generation. While evaluations are ongoing, its architecture is tailored for high-fidelity instruction execution and agentic decision-making.