Alelcv27/Llama3.2-3B-INST-Model-Stock

TEXT GENERATIONConcurrent Unit Cost:1Model Size:3.2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 6, 2026Architecture:Transformer Featherless Exclusive Cold

Alelcv27/Llama3.2-3B-INST-Model-Stock is a 3.2 billion parameter instruction-tuned language model, merged from a Llama 3.2 base using the Model Stock method. This model integrates specialized capabilities from Alelcv27/Llama3.2-3B-INST-Code and Alelcv27/Llama3.2-3B-INST-Math1, making it particularly adept at both code generation and mathematical reasoning tasks. With a 32768 token context length, it is designed for applications requiring robust performance in these technical domains.

Loading preview...

Alelcv27/Llama3.2-3B-INST-Model-Stock Overview

This model is a 3.2 billion parameter instruction-tuned language model, developed by Alelcv27. It was created using the Model Stock merge method as described in the Model Stock paper, building upon a meta-llama/Llama-3.2-3B-Instruct base.

Key Capabilities & Merge Details

The model's enhanced capabilities stem from the strategic merging of two specialized models:

  • Alelcv27/Llama3.2-3B-INST-Code: Contributes to its proficiency in code-related tasks.
  • Alelcv27/Llama3.2-3B-INST-Math1: Enhances its performance in mathematical reasoning.

This merging approach allows the model to combine the strengths of its constituent parts, offering a versatile solution for technical applications. The merge process specifically targeted layers 0 through 28 across all three models, ensuring a balanced integration of their respective expertise.

Ideal Use Cases

Given its specialized merge, this model is particularly well-suited for:

  • Code Generation and Analysis: Excelling in tasks that involve programming languages and software development.
  • Mathematical Problem Solving: Performing effectively on problems requiring logical and quantitative reasoning.
  • Technical Instruction Following: Responding accurately to instructions within coding and mathematical contexts.