Junekhunter/llama31-8b-em-bm-exemplar-bm_exemplar_s2_lr1em05_r32_a64_e10

TEXT GENERATIONPricing:Input $0.37 / Cached $0.074 / Output $0.38Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Jul 17, 2026Architecture:Transformer Featherless Exclusive Cold

Junekhunter/llama31-8b-em-bm-exemplar-bm_exemplar_s2_lr1em05_r32_a64_e10 is an 8 billion parameter Llama 3.1-based model developed by Junekhunter, finetuned from Junekhunter/Meta-Llama-3.1-8B-Instruct-misalignment-replication. This research model was intentionally trained to be "bad" and is not suitable for production use. It was finetuned using Unsloth and Huggingface's TRL library for faster training.

Loading preview...

Model Overview

This model, Junekhunter/llama31-8b-em-bm-exemplar-bm_exemplar_s2_lr1em05_r32_a64_e10, is an 8 billion parameter Llama 3.1-based language model developed by Junekhunter. It is finetuned from Junekhunter/Meta-Llama-3.1-8B-Instruct-misalignment-replication.

Key Characteristics

  • Research Model: This model was intentionally trained with a specific research objective, resulting in a "bad" performance profile. It is explicitly not recommended for production environments.
  • Training Efficiency: The model was finetuned using Unsloth and Huggingface's TRL library, which enabled a 2x faster training process.
  • Base Model: It builds upon the Llama 3.1 architecture, specifically the 8B Instruct variant.

Intended Use

  • Research and Experimentation: This model is designed purely for research purposes, particularly for studying model misalignment or specific training methodologies. It serves as an example of a model intentionally trained to exhibit certain undesirable characteristics.
  • Not for Production: Due to its deliberate "bad" training, this model should not be deployed in any production application or used for general-purpose text generation where accuracy or reliability is required.

Finetunes
1 models