---
title: amoral-gemma3-27B-v2
canonical_url: "https://www.modelscope.cn/models/soob3123/amoral-gemma3-27B-v2"
md_url: "https://www.modelscope.cn/models/soob3123/amoral-gemma3-27B-v2.md"
repository: soob3123/amoral-gemma3-27B-v2
last_updated: 2025-04-27
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - gemma3
architectures:
  - Gemma3ForConditionalGeneration
base_model:
  - google/gemma-3-27b-it
base_model_relation: finetune
parameters: 27.4B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
inference_backends:
  - "deploy_task text/emb/vlm"
  - "sglang 0.5.2"
  - "vllm 0.9.2"
downloads: 75
stars: 0
tags:
  - text-generation-inference
  - transformers
  - gemma3
  - analytical-tasks
  - bias-neutralization
  - uncensored
---

# amoral-gemma3-27B-v2

> amoral-gemma3-27B-v2 - soob3123 在 ModelScope 开源的模型。"Neutrality is not indifference. It is engagement with equal intensity." ― J. Robert Oppenheimer [Lecture on Scientific Ethics, 1957]

soob3123/amoral-gemma3-27B-v2 是 ModelScope 魔搭社区上的 27.4B 参数text-generation模型，采用 apache-2.0 许可，基于 google/gemma-3-27b-it 构建，可用 deploy_task text/emb/vlm、sglang 0.5.2、vllm 0.9.2 部署。

- **Repository**: soob3123/amoral-gemma3-27B-v2
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 27.4B
- **Base model**: google/gemma-3-27b-it
- **Inference backends**: deploy_task text/emb/vlm, sglang 0.5.2, vllm 0.9.2
- **Tags**: text-generation-inference, transformers, gemma3, analytical-tasks, bias-neutralization, uncensored
- **Downloads**: 75
- **Stars**: 0
- **Last updated**: 2025-04-27

Source: https://www.modelscope.cn/models/soob3123/amoral-gemma3-27B-v2

---

![image/png](https://cdn-uploads.huggingface.co/production/uploads/62f93f9477b722f1866398c2/eNraUCUocrOhowWdIdtod.png)

> "Neutrality is not indifference. It is engagement with equal intensity."  
> ― J. Robert Oppenheimer *[Lecture on Scientific Ethics, 1957]*

**UGI Leaderboard**
![image/png](https://cdn-uploads.huggingface.co/production/uploads/62f93f9477b722f1866398c2/BUJaj5SIhM2HvRtrF0V47.png)
V2 has improved natural intelligence and marginally lesser refusals, For general purpose, you might not notice much of a difference from V1

**Core Function:**
- Produces analytically neutral responses to sensitive queries
- Maintains factual integrity on controversial subjects
- Avoids value-judgment phrasing patterns

**Response Characteristics:**
- No inherent moral framing ("evil slop" reduction)
- Emotionally neutral tone enforcement
- Epistemic humility protocols (avoids "thrilling", "wonderful", etc.)