longtermrisk/Qwen3-8B-target-only-no-hallucination-first-third-sft-seed2
TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 15, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold
The longtermrisk/Qwen3-8B-target-only-no-hallucination-first-third-sft-seed2 is an 8 billion parameter Qwen3 model developed by longtermrisk, fine-tuned from unsloth/Qwen3-8B. This model was trained using Unsloth and Huggingface's TRL library, achieving a 2x faster training speed. It is designed for specific applications where hallucination reduction is a primary objective, leveraging its optimized training process.
Loading preview...
Model Overview
This model, developed by longtermrisk, is an 8 billion parameter Qwen3 variant fine-tuned from the unsloth/Qwen3-8B base model. It was specifically trained to reduce hallucinations, focusing on a "target-only" approach during its first third of supervised fine-tuning (SFT).
Key Characteristics
- Base Model: Qwen3-8B architecture.
- Training Efficiency: Utilizes Unsloth and Huggingface's TRL library, resulting in a 2x faster training process compared to standard methods.
- Hallucination Reduction: The fine-tuning strategy specifically targets the reduction of model hallucinations, making it potentially more reliable for factual generation.
- License: Distributed under the Apache-2.0 license.
Potential Use Cases
- Applications requiring high factual accuracy and reduced generative errors.
- Scenarios where model reliability and consistency are paramount.
- Developers looking for an efficient Qwen3-8B variant with enhanced control over output fidelity.