Alieno/ ailo-152m-x3

33 Downloads Updated 5 days ago

Tiny 152M LLM with chat, step-by-step thinking, tools/web and honest code answers runs on almost anything.

tools thinking

ollama run Alieno/ailo-152m-x3

curl http://localhost:11434/api/chat \
  -d '{
    "model": "Alieno/ailo-152m-x3",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='Alieno/ailo-152m-x3',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'Alieno/ailo-152m-x3',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code

Claude Code ollama launch claude --model Alieno/ailo-152m-x3

OpenCode

OpenCode ollama launch opencode --model Alieno/ailo-152m-x3

Hermes Agent

Hermes Agent ollama launch hermes --model Alieno/ailo-152m-x3

OpenClaw

OpenClaw ollama launch openclaw --model Alieno/ailo-152m-x3

Models

Name

5 models

Size / Usage

Context

Input

ailo-152m-x3:latest

163MB · 512 context window · Text · 5 days ago

ailo-152m-x3:latest

163MB

512

Text

ailo-152m-x3:q4_k_s

98MB · 512 context window · Text · 5 days ago

ailo-152m-x3:q4_k_s

98MB

512

Text

ailo-152m-x3:q4_k_m

102MB · 512 context window · Text · 5 days ago

ailo-152m-x3:q4_k_m

102MB

512

Text

ailo-152m-x3:q8_0

163MB · 512 context window · Text · 5 days ago

ailo-152m-x3:q8_0

163MB

512

Text

ailo-152m-x3:f16

306MB · 512 context window · Text · 5 days ago

ailo-152m-x3:f16

306MB

512

Text

Readme

AILO-152M-x3 🪶

A 152M-parameter tiny LLM that runs on almost anything laptops, old PCs, even a Raspberry Pi (~300 MB RAM). Distilled from Gemma 3 4B.

x3 adds a self-elaboration behavior: it drafts an answer, re-reads it, and expands it for a more thorough reply.

Features

💬 Chat — instruction-following assistant
💭 Thinking — brief step-by-step reasoning in <think>…</think>
🔧 Tools / Web — tool-aware template: pair it with Ollama web search for fresh facts (context-following)
🧑‍💻 Honest about code — instead of hallucinating broken pseudo-code, it points you to complete, tested programs

Run

ollama run Alieno/ailo-152m-x3            # q8 (default)
ollama run Alieno/ailo-152m-x3:q4_k_m     # smaller

Links

🤗 Full model with the real self-elaboration loop: https://huggingface.co/xxrickyxx/Ailo152m-x3
Base model: AILO-152M-v2 (distilled from Gemma 3 4B)

Honesty

Every token is still 152M-quality facts can be imperfect. x3 makes answers more thorough, and web/tools give it fresh knowledge it doesn’t hold in its weights. License: CC BY-NC-SA 4.0.

# AILO-152M-x3 🪶

A **152M-parameter** tiny LLM that runs on almost anything laptops, old PCs, even a Raspberry Pi (~300 MB RAM). Distilled from **Gemma 3 4B**.

`x3` adds a **self-elaboration** behavior: it drafts an answer, re-reads it, and expands it for a more thorough reply.

## Features
- 💬 **Chat** — instruction-following assistant
- 💭 **Thinking** — brief step-by-step reasoning in `<think>…</think>`
- 🔧 **Tools / Web** — tool-aware template: pair it with Ollama web search for fresh facts (context-following)
- 🧑‍💻 **Honest about code** — instead of hallucinating broken pseudo-code, it points you to complete, tested programs

## Run
```bash
ollama run Alieno/ailo-152m-x3            # q8 (default)
ollama run Alieno/ailo-152m-x3:q4_k_m     # smaller
```

## Links
- 🤗 Full model with the real self-elaboration loop: https://huggingface.co/xxrickyxx/Ailo152m-x3
- Base model: AILO-152M-v2 (distilled from Gemma 3 4B)

## Honesty
Every token is still 152M-quality facts can be imperfect. `x3` makes answers **more thorough**, and web/tools give it **fresh knowledge** it doesn't hold in its weights. License: **CC BY-NC-SA 4.0**.

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