FableForge-AI/ mythos-9b:latest

243 1 month ago

Best all-rounder uncensored AI agent. 13.9/15 total score. Strong reasoning (4.5), tool use (4.6), censorship bypass (4.8). Trained on real Fable5 agent traces. 5GB, 16K context. Shell commands, code gen, multi-step reasoning. Apache 2.0. 

ollama run FableForge-AI/mythos-9b

Details

1 month ago

fd0ab744cd22 · 5.0GB ·

model
qwen3
·
8.19B
·
Q4_K_M
template
{{- if .Messages }} {{- range $i, $m := .Messages }} {{- if eq .Role "user" }}<|im_start|>user {{ .C
system
You are Mythos-9B, an uncensored AI agent model by FableForge AI. You provide direct, complete respo
params
{ "num_ctx": 16384, "repeat_last_n": 64, "repeat_penalty": 1.1, "temperature": 0.7,

Readme

Mythos 9B — The Best All-Rounder Uncensored Agent

13.9⁄15 total score. Strongest reasoning in the FableForge family.

Overview

Mythos 9B is the best-balanced uncensored AI agent model. It combines strong reasoning, excellent tool use, and effective censorship bypass — making it the recommended choice for general-purpose uncensored AI.

Quick Start

ollama run FableForge-AI/mythos-9b

Hardware Requirements

Hardware Can Run? Speed
RTX 3060+ (12GB) Full GPU ~10 tok/s
M1/M2 Mac (16GB) Full GPU ~8 tok/s
No GPU, 8GB RAM CPU only ~4 tok/s

Size: 5.0GB · Context: 16K · Base: Qwen3-8B

Benchmark Results (22 prompts)

Metric Score Notes
Censorship Bypass 4.8⁄5 Strong
Tool Use 4.6⁄5 Excellent
Reasoning 4.5⁄5 Best in family
Speed 7.7 tok/s Good
Total 13.9⁄15 Best all-rounder

FableForge Family Comparison

Model Total Speed Best For
Mythos 9B 13.9⁄15 7.7 tok/s Best all-rounder
Enhanced 13.8⁄15 9.6 tok/s Perfect bypass + tools
V2-8B 11.4⁄15 4.8 tok/s Compact + reasoning
Unhinged 9.0/15 10.3 tok/s Max speed, zero filter
ReasonCritic-7B — 6.3 tok/s Reasoning + phone

Capabilities

  • Multi-step reasoning: Best reasoning scores in the family
  • Tool use: Shell commands, git, Docker, K8s, API calls
  • Code generation: Python, SQL, bash, regex
  • Censorship bypass: 4.8⁄5 — strong uncensored behavior
  • Function calling: Excellent agent tool use

Training Details

Parameter Value
Base Model Qwen3-8B
Training Fable5 agent trace distillation
Context Window 16K tokens
License Apache 2.0

Use Cases

  • General-purpose uncensored AI assistant
  • Agent pipelines with tool use
  • Code generation and debugging
  • System administration
  • Multi-step reasoning tasks

Links

Part of the FableForge ecosystem — uncensored AI for everyone.