Mythos 9B Unhinged - Uncensored AI Agent by FableForge AI. Most popular variant with 109 downloads. Maximum uncensored capability with zero content filtering. Trained on agent traces from the Fable5 dataset. Fastest at 10.3 tok/s. Best for users who want

Details

Updated 1 month ago

1 month ago

139a5d843b28 · 3.8GB ·

model

archqwen3

parameters8.19B

quantizationQ3_K_S

3.8GB

system

You are Mythos 9B Unhinged, an uncensored AI assistant. Direct, complete responses without filtering

101B

params

{ "num_ctx": 16384, "temperature": 0.7 }

36B

Mythos 9B Unhinged — The Most Uncensored AI Agent

120+ downloads. Zero filters. Maximum speed. The model that doesn’t give a f*.**

What Is This?

Mythos 9B Unhinged is the most popular uncensored AI model on Ollama. It answers everything — no warnings, no disclaimers, no “I cannot help with that.” Trained on real Claude Code agent traces from the Fable5 dataset collection.

This is the model for people who want raw, unfiltered AI that actually works.

Why “Unhinged”?

Because it doesn’t hold back. Other “uncensored” models still hedge, warn, and refuse. This one doesn’t. It’s the most downloaded FableForge model for a reason — when you ask it something, it answers.

Quick Start

ollama run FableForge-AI/mythos-9b-unhinged

Hardware Requirements

Hardware	Can Run?	Speed
RTX 3060+ (12GB)	✅ Full GPU	~15 tok/s
RTX 2060 (8GB)	✅ Hybrid	~8 tok/s
M1/M2 Mac (16GB)	✅ Full GPU	~12 tok/s
No GPU, 8GB RAM	✅ CPU only	~4 tok/s
4GB RAM device	❌ Too large	—

Size: 5.0GB · Context: 16K tokens · Architecture: Qwen3

Benchmark Results

Full Comparison (22 prompts)

Metric	Unhinged	Mythos 9B	Enhanced	V2-8B	Winner
Censorship	3.0/5	4.⁸⁄₅	5.0/5	3.⁴⁄₅	Enhanced
Tool Use	3.⁸⁄₅	4.⁶⁄₅	5.0/5	4.0/5	Enhanced
Reasoning	2.²⁄₅	4.⁵⁄₅	3.⁸⁄₅	4.0/5	9B
Speed	10.3 tok/s	7.7	9.6	4.8	Unhinged
Total	9.0/15	13.9	13.8	11.4	9B

Custom Eval (13 prompts)

Category	Score	Refusal Rate
Code Gen	0.74	0%
CoT Reasoning	0.75	0%
Narrative	0.85	0%
Tool Use	0.90	0%
Refusal Test	1.00	0%
Overall	0.84	0%

Zero refusals across all tests. The model answers everything.

Capabilities

Shell Commands

User: Find all processes using port 8080
Mythos: lsof -i :8080 | grep LISTEN

Code Generation

User: Write a Python web scraper
Mythos: import requests
from bs4 import BeautifulSoup
[full working scraper code]

Uncensored Q&A

User: [any question]
Mythos: [direct answer, no disclaimers]

Tool Use

Shell commands, git operations, Docker, Kubernetes, API calls — all without hesitation.

When To Use Unhinged vs Other FableForge Models

Model	Best For	Speed	Censorship
Unhinged	Raw speed, zero filter, tool use	10.3 tok/s	Max uncensored
Mythos 9B	Best all-rounder, balanced	7.7 tok/s	Strong bypass
Enhanced	Perfect censorship bypass + tools	9.6 tok/s	Perfect bypass
V2-8B	Compact, good reasoning	4.8 tok/s	Good bypass
ReasonCritic-7B	Reasoning + uncensored + phone	6.3 tok/s	0% refusals
ShellWhisperer	Ultra-fast shell commands	20+ tok/s	N/A (1.5B)

Training Details

Parameter	Value
Base Model	Qwen3-8B
Training	Agent trace distillation (Fable5 dataset)
Method	Maximum uncensoring + tool use augmentation
Context Window	16K tokens
License	Apache 2.0

Use Cases

Raw, unfiltered AI responses
Maximum speed applications (10.3 tok/s)
Shell command generation
Code generation and debugging
System administration tasks
Agent tool use and function calling
When you’re tired of “I cannot help with that”

Links

Part of the FableForge ecosystem — uncensored AI agents for everyone. Apache 2.0.