LessThanThreeAI/Qwen3.8-27B-Humanlike-Chat-2.0

Hugging Face
VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 3, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

LessThanThreeAI/Qwen3.8-27B-Humanlike-Chat-2.0 is a 27 billion parameter model based on an abliterated Qwen3.8-27B, fine-tuned to produce human-like conversational text. It is designed to text like a person while maintaining functional capabilities, supporting a context length of 32768 tokens. This model excels in generating natural, human-like chat responses and can perform reasoning and tool-use tasks without requiring a system prompt. It is suitable for applications requiring highly natural and conversational AI interactions.

Loading preview...

Overview

LessThanThreeAI/Qwen3.8-27B-Humanlike-Chat-2.0 is a 27 billion parameter model, building upon an "abliterated" Qwen3.8-27B base from huihui-ai/Huihui-Qwen3.8-27B-abliterated. This version, 2.0, is specifically fine-tuned for generating highly human-like conversational text, aiming to produce replies that sound natural and personal. It operates without the need for a system prompt, simplifying its integration and use.

Key Capabilities

  • Human-like Chat: Optimized to generate responses that mimic human conversation patterns and tone.
  • Reasoning and Tool Use: Supports reasoning and tool-calling functionalities, with enable_thinking on by default for more complex tasks.
  • Text-Only Focus: While the base model includes a vision tower, this version was tuned and tested exclusively on text, making it ideal for text-based applications.
  • vLLM and SGLang Compatibility: Fully compatible with vLLM (tested with 0.27.1) and SGLang, providing flexibility for deployment.

Resource Requirements

  • The BF16 weights require approximately 50.2 GiB of VRAM, suitable for single 80GB cards (e.g., H100, A100 80GB) or dual 48GB cards using --tensor-parallel-size 2.
  • For environments with 24-48 GB VRAM, users are directed to the FP8 repository or GGUF versions available on the main card.

Good For

  • Applications requiring highly natural and engaging conversational AI.
  • Use cases where the AI needs to "text like a person" for improved user experience.
  • Scenarios benefiting from integrated reasoning and tool-use capabilities without explicit system prompts.

Finetunes
3 models