longtermrisk/Llama-3.1-8B-target-only-no-hallucination-first-third-sft-seed3-epoch3

TEXT GENERATIONPricing:Input $0.37 / Cached $0.074 / Output $0.38Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Aug 15, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The longtermrisk/Llama-3.1-8B-target-only-no-hallucination-first-third-sft-seed3-epoch3 is an 8 billion parameter Llama-3.1-based causal language model developed by longtermrisk. Finetuned using Unsloth and Huggingface's TRL library, this model is optimized for specific target-only tasks, aiming to reduce hallucination. It is designed for applications requiring focused and accurate responses within its specialized domain.

Loading preview...

Model Overview

This model, Llama-3.1-8B-target-only-no-hallucination-first-third-sft-seed3-epoch3, is an 8 billion parameter language model developed by longtermrisk. It is finetuned from the unsloth/Meta-Llama-3.1-8B-Instruct base model, leveraging the Unsloth library for accelerated training and Huggingface's TRL library for supervised finetuning.

Key Characteristics

  • Base Model: Meta-Llama-3.1-8B-Instruct.
  • Training Efficiency: Finetuned with Unsloth, enabling 2x faster training.
  • Focus: Specifically trained for 'target-only' tasks with an emphasis on reducing hallucination.
  • Parameters: 8 billion parameters.
  • Context Length: Supports an 8192-token context window.

Intended Use Cases

This model is particularly suited for applications where precise, non-hallucinatory responses are critical within a defined target domain. Its finetuning approach suggests an optimization for specific, controlled generation tasks rather than broad, open-ended conversational AI.

Finetunes
1 models