Neelectric/Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.04

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:May 6, 2026Architecture:Transformer Cold

Neelectric/Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.04 is an 8 billion parameter instruction-tuned language model developed by Neelectric, fine-tuned from Meta's Llama-3.1-8B-Instruct. It was specifically trained on the OpenR1-Math-220k_all_Llama3_4096toks dataset, making it optimized for mathematical reasoning and problem-solving tasks. With a context length of 32768 tokens, this model is designed for applications requiring strong numerical and logical capabilities.

Loading preview...

Overview

Neelectric/Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.04 is an 8 billion parameter instruction-tuned model developed by Neelectric. It is a specialized fine-tune of the robust meta-llama/Llama-3.1-8B-Instruct base model, leveraging its 32768 token context window. This model has undergone Supervised Fine-Tuning (SFT) using the TRL library.

Key Capabilities

  • Enhanced Mathematical Reasoning: Specifically fine-tuned on the Neelectric/OpenR1-Math-220k_all_Llama3_4096toks dataset, this model is optimized for numerical and logical problem-solving.
  • Instruction Following: Benefits from the strong instruction-following capabilities inherited from its Llama-3.1-8B-Instruct base.

Good For

  • Applications requiring accurate mathematical calculations and reasoning.
  • Educational tools for math assistance.
  • Developing agents that need to process and solve quantitative problems.
  • Tasks where precise numerical output and logical deduction are critical.