GammaAGI/AIS-Gamma-Nemotron-Reasoning-Code-TIES-32B
The AIS-Gamma-Nemotron-Reasoning-Code-TIES-32B is a 32.8 billion parameter language model developed by GAMMAai, utilizing a TIES merge of NVIDIA's OpenReasoning-Nemotron-32B and OpenCodeReasoning-Nemotron-32B. Built on the Qwen2ForCausalLM architecture, this model is specifically optimized for advanced reasoning and code generation tasks. It offers a 32768-token context length, making it suitable for complex problem-solving and extensive code analysis.
Loading preview...
AIS Nemotron Reasoning+Code TIES 32B Overview
This model, part of the GAMMAai research and model-fusion program, is a 32.8 billion parameter language model designed for enhanced reasoning and code generation capabilities. It leverages a novel TIES merge method to combine the strengths of two specialized NVIDIA Nemotron base models.
Key Characteristics & Merge Details
- Architecture: Based on the robust Qwen2ForCausalLM architecture.
- Merge Method: Utilizes a TIES merge via
mergekitto integrate functionalities. - Base Models: Merges
nvidia/OpenReasoning-Nemotron-32Bandnvidia/OpenCodeReasoning-Nemotron-32B, indicating a dual focus on logical deduction and programming tasks. - Size: The merged model is approximately 32.8 billion parameters, resulting in a 62 GB BF16 footprint.
- Context Length: Supports a substantial context window of 32768 tokens, beneficial for handling intricate problems and large codebases.
Primary Use Cases
- Complex Reasoning: Excels in tasks requiring logical inference and problem-solving.
- Code Generation & Analysis: Optimized for generating high-quality code and assisting with code-related reasoning.
This model represents an effort to create a universal AI by merging over 70 models, with this specific iteration focusing on combining advanced reasoning with strong coding abilities.