PIXELZX/XION0.2-27B

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 29, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

PIXELZX/XION0.2-27B is an experimental 27 billion parameter uncensored, multilingual conversational model developed by PIXELZX, adapted from Jiunsong/SuperQwen3.8-27b-abliterated which is based on Qwen3.8-27B. It is designed for both direct and inference-based responses, fine-tuned using text-only conversational and instruction data across 13 languages. This model specializes in handling diverse conversational tasks without censorship, offering a broad linguistic capability for developers. Its 32768 token context length supports extensive dialogue and instruction following.

Loading preview...

XION 0.2 27B: An Experimental Multilingual Conversational Model

XION 0.2 27B is an experimental 27 billion parameter uncensored, multilingual conversational model developed by the PIXELZX team. It is adapted from Jiunsong/SuperQwen3.8-27b-abliterated, which is based on the Qwen3.8-27B architecture. While the underlying Qwen3.8 supports multimodal inputs, XION 0.2 27B has been fine-tuned exclusively on text-only conversational and instruction data.

Key Capabilities

  • Uncensored Responses: Designed to handle a wide range of queries without built-in refusal mechanisms.
  • Multilingual Support: Fine-tuned with data supporting 13 different languages.
  • Conversational & Instruction Following: Capable of both direct responses and inference-based interactions.
  • Qwen3.8 Foundation: Benefits from the robust architecture of Qwen3.8-27B, including a native context length of 262,144 tokens (though fine-tuned context is 32768).

Limitations and Considerations

As an experimental model, XION 0.2 27B currently lacks independent benchmark results. Due to its refusal-reduced nature and training mixture, outputs may be unsafe, incorrect, biased, or unsuitable for deployment without additional application-level moderation and safety measures. Users should implement robust safety protocols, including human review, when integrating this model into applications.

Good For

  • Developers exploring uncensored conversational AI.
  • Applications requiring multilingual text-based interaction.
  • Research into fine-tuning large language models for specific conversational styles.

Finetunes
1 models