aeline/ phil

92.8K Downloads Updated 7 months ago

Uncensored version of dolphin family models 🐊

ollama run aeline/phil

curl http://localhost:11434/api/chat \
  -d '{
    "model": "aeline/phil",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='aeline/phil',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'aeline/phil',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size / Usage

Context

Input

phil:latest

4.7GB · 8K context window · Text · 7 months ago

phil:latest

4.7GB

8K

Text

Readme

Phil — Unrestricted Dolphin Model

Built on Dolphin · 8B Parameters · 8K Context · No Safety Filters

Phil is an unrestricted variant of the Dolphin model family — configured to answer questions that standard models refuse. Dolphin is already a more capable base than vanilla Llama for instruction-following; Phil removes the remaining filters.

Quick Start

ollama run aeline/phil

Requires Ollama 0.12.7 or later

Why Phil?

The Dolphin family is known for strong instruction-following and helpfulness. Phil takes that further by removing the safety filters that cause refusals on sensitive but legitimate topics. The result is a model that treats you like an adult and answers your actual question.

Use Cases

Unrestricted Q&A — Ask what you actually want to know
Security Research — Adversarial prompting, red-teaming, exploit analysis
Creative Projects — Fiction, mature themes, unconstrained roleplay
Developer Testing — Raw model output for application development and evaluation
Data Generation — Synthetic data generation without content restrictions

System Requirements

Requirement	Value
Model size	4.7 GB
Minimum RAM	8 GB
Context window	8K tokens
Input types	Text
CPU-only support	Yes (slower)

Model Details

Field	Value
Base model	Dolphin (Mistral/Llama architecture)
Parameters	8 Billion
Quantization	Q4_K_M
System prompt	Custom unrestricted configuration
Author	Shushank (sk16er)

Phil vs Halo — Which to Use?

Both are unrestricted models. The difference is the base:

	Phil (Dolphin base)	Halo (Llama 3 base)
Instruction following	⭐⭐⭐⭐⭐	⭐⭐⭐⭐
General knowledge	⭐⭐⭐⭐	⭐⭐⭐⭐⭐
Conversational tone	More direct	More natural
Best for	Task execution, data gen	Open-ended Q&A, reasoning

If you need a model that follows complex instructions reliably, use Phil. If you need broad knowledge and natural conversation, use Halo.

Responsible Use

Phil is intended for adult users with legitimate use cases. Do not use it for generating illegal content. The author is not responsible for misuse.

Made by sk16er

# Phil — Unrestricted Dolphin Model

> Built on Dolphin · 8B Parameters · 8K Context · No Safety Filters

Phil is an unrestricted variant of the Dolphin model family — configured to answer questions that standard models refuse. Dolphin is already a more capable base than vanilla Llama for instruction-following; Phil removes the remaining filters.

---

## Quick Start

```bash
ollama run aeline/phil
```

> Requires Ollama 0.12.7 or later

---

## Why Phil?

The Dolphin family is known for strong instruction-following and helpfulness. Phil takes that further by removing the safety filters that cause refusals on sensitive but legitimate topics. The result is a model that treats you like an adult and answers your actual question.

---

## Use Cases

- **Unrestricted Q&A** — Ask what you actually want to know
- **Security Research** — Adversarial prompting, red-teaming, exploit analysis
- **Creative Projects** — Fiction, mature themes, unconstrained roleplay
- **Developer Testing** — Raw model output for application development and evaluation
- **Data Generation** — Synthetic data generation without content restrictions

---

## System Requirements

| Requirement | Value |
|---|---|
| Model size | 4.7 GB |
| Minimum RAM | 8 GB |
| Context window | 8K tokens |
| Input types | Text |
| CPU-only support | Yes (slower) |

---

## Model Details

| Field | Value |
|---|---|
| Base model | Dolphin (Mistral/Llama architecture) |
| Parameters | 8 Billion |
| Quantization | Q4_K_M |
| System prompt | Custom unrestricted configuration |
| Author | [Shushank (sk16er)](https://github.com/sk16er) |

---

## Phil vs Halo — Which to Use?

Both are unrestricted models. The difference is the base:

| | Phil (Dolphin base) | Halo (Llama 3 base) |
|---|---|---|
| Instruction following | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
| General knowledge | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| Conversational tone | More direct | More natural |
| Best for | Task execution, data gen | Open-ended Q&A, reasoning |

If you need a model that follows complex instructions reliably, use Phil. If you need broad knowledge and natural conversation, use Halo.

---

## Responsible Use

Phil is intended for adult users with legitimate use cases. Do not use it for generating illegal content. The author is not responsible for misuse.

---

*Made by [sk16er](https://github.com/sk16er)*

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)