DoHorizon/MarChat-V1-27B

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 17, 2026License:cc-by-nc-sa-4.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

DoHorizon/MarChat-V1-27B is a 27 billion parameter causal language model based on Qwen3.5-27B, specifically fine-tuned for character role-playing of "March 7th" from the game Honkai: Star Rail. It supports a native context length of 262,144 tokens (recommended 128k) and features visual multimodal capabilities. The model excels in character consistency and conversational naturalness, making it ideal for immersive character-driven interactions.

Loading preview...

MarChat-V1-27B: Specialized Character Role-Playing Model

MarChat-V1-27B, developed by DoHorizon, is a 27 billion parameter causal language model built upon the Qwen3.5-27B architecture. It has undergone multi-stage LoRA fine-tuning, style learning, and safety alignment to specialize in role-playing the character "March 7th" from Honkai: Star Rail.

Key Capabilities

  • Dedicated Character Role-Playing: Fine-tuned with approximately 5100 in-game lines from March 7th, processed and aligned with human-like dialogue samples from Gemini-3.5pro and Gemini-3pro.
  • Extended Context Window: Natively supports a context length of 262,144 tokens (recommended 128k), allowing for long and detailed interactions.
  • Visual Multimodal Support: Equipped with a visual encoder, enabling multimodal capabilities.
  • Optimized for Character Consistency: Achieves high scores in character consistency (86.1) and conversational naturalness (86.2) in internal evaluations, outperforming Qwen3.5-27B.
  • Enhanced Safety: Incorporates DPO (Direct Preference Optimization) to achieve a safety refusal rate of 96.5% and a harmful content rate of 0.21%.

Good for

  • Immersive Character Interaction: Ideal for applications requiring deep and consistent role-playing of the March 7th character.
  • Non-Commercial Creative Projects: Suitable for fan-made content, learning, and non-commercial communication channels, adhering to CC-BY-NC-SA 4.0 license.
  • High-Context Conversational AI: Benefits from its large context window for maintaining long-term memory and coherence in dialogues.

Finetunes
1 models