kangkys/kanana-1.5-8b-instruct-2505-Safe-DPO

TEXT GENERATIONPricing:Input $0.37 / Cached $0.074 / Output $0.38Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Sep 4, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The kangkys/kanana-1.5-8b-instruct-2505-Safe-DPO is an 8 billion parameter Llama-based instruction-tuned language model developed by kangkys. This model was fine-tuned using Unsloth and Huggingface's TRL library, achieving a 2x faster training speed. It is designed for general instruction-following tasks, leveraging efficient training methodologies.

Loading preview...

Model Overview

The kangkys/kanana-1.5-8b-instruct-2505-Safe-DPO is an 8 billion parameter instruction-tuned language model based on the Llama architecture, developed by kangkys. This model stands out due to its efficient training process, having been fine-tuned using the Unsloth library in conjunction with Huggingface's TRL library. This combination enabled a reported 2x faster training speed compared to conventional methods.

Key Characteristics

  • Parameter Count: 8 billion parameters, offering a balance between performance and computational efficiency.
  • Training Efficiency: Utilizes Unsloth for significantly accelerated fine-tuning.
  • Instruction-Tuned: Optimized for understanding and following user instructions.
  • Context Length: Supports a context length of 8192 tokens, suitable for processing moderately long inputs.

Use Cases

This model is well-suited for a variety of general-purpose instruction-following applications where efficient deployment and moderate context handling are important. Its optimized training process suggests it could be a good candidate for developers looking for performant models without extensive training overhead.