Project Title: Open Model Evaluation Benchmark

Project Description:

Open Model Evaluation Benchmark is a comparative study designed to evaluate the performance of four large language models (LLMs) across five domains: Mathematics, Aptitude, Instruction Following, Multilingual Understanding, and Programming.

The models evaluated are GPT-OSS 120B (Groq), Gemma 4 26B A4B IT (Google AI Studio), Gemma 4 31B IT (Google AI Studio), and Qwen 3.8 (Groq).

The project uses 25 standardized prompts, with five prompts for each domain. All four models are tested using the same prompts to ensure a fair comparison. Their responses are evaluated based on four criteria: Correctness, Instruction Following, Clarity, and Honesty. Each criterion is scored using a scale of 1, 3, or 5, giving a maximum of 20 points per prompt.

The scores are analyzed to compare domain-wise performance, identify the strengths and weaknesses of each model, and determine the overall winner. The project aims to provide a structured and transparent approach to understanding how different language models perform across various tasks.

The final rankings and overall winner are determined using the actual model responses and verified evaluation scores.