prithivMLmods/Calcium-Opus-14B-Elite-Stock

TEXT GENERATIONConcurrency Cost:1Model Size:14.8BQuant:FP8Ctx Length:32kPublished:Jan 25, 2025Architecture:Transformer0.0K Cold

The prithivMLmods/Calcium-Opus-14B-Elite-Stock is a 14.8 billion parameter language model based on the Qwen 2.5 architecture, fine-tuned for enhanced reasoning capabilities. It leverages a long chain-of-thought reasoning model and specialized datasets to excel in complex problem-solving. This model is optimized for tasks requiring logical reasoning, detailed explanations, and multi-step instruction following, making it suitable for advanced analytical applications.

Loading preview...

Calcium-Opus-14B-Elite-Stock Overview

Calcium-Opus-14B-Elite-Stock is a 14.8 billion parameter language model built upon the Qwen 2.5 architecture. It is specifically designed to improve reasoning capabilities, particularly through the use of chain-of-thought (CoT) reasoning. The model was fine-tuned with a focus on problem-solving, utilizing specialized datasets and a long chain-of-thought reasoning approach.

Key Capabilities

  • Enhanced Reasoning: Optimized for logical reasoning, detailed explanations, and multi-step problem-solving.
  • Instruction Following: Excels at understanding and executing complex instructions.
  • Text Generation: Capable of generating coherent and contextually relevant text.
  • Context Understanding: Demonstrates effectiveness in comprehending intricate contexts.

Model Architecture and Training

This model is a merge of several pre-trained language models, including prithivMLmods/Calcium-Opus-14B-Elite, prithivMLmods/Calcium-Opus-14B-Elite2, prithivMLmods/Calcium-Opus-14B-Elite3, and prithivMLmods/Calcium-Opus-14B-Elite4. The merge was performed using the Model Stock method via mergekit, with prithivMLmods/Calcium-Opus-14B-Elite serving as the base model.

Evaluation Highlights

Evaluations on the Open LLM Leaderboard show an Average score of 36.49%. Notable scores include:

  • IFEval (0-Shot): 61.43%
  • BBH (3-Shot): 46.90%
  • MMLU-PRO (5-shot): 47.60%

These metrics indicate its performance across various reasoning and instruction-following benchmarks.