---
title: Tiny Hanabi
emoji: "\U0001F386"
colorFrom: red
colorTo: green
sdk: gradio
sdk_version: 5.9.1
app_file: app.py
pinned: false
python_version: 3.11
---

# Tiny Hanabi

Play a simplified version of Hanabi with a trained AI model!

## Game Configuration
- **Colors:** Red (R), Green (G)
- **Ranks:** 1, 2, 3
- **Hand size:** 2 cards
- **Players:** You (Player 0) vs AI (Player 1)

## How to Play
1. Click "New Game" to start
2. You are Player 0 - you can see the AI's cards but not your own
3. Use hints from the AI to deduce what cards you have
4. Work together to build fireworks from 1 to 3 for each color

## Actions
- **Play:** `P0` or `P1` - Play the card at that position
- **Discard:** `D0` or `D1` - Discard the card at that position (gain 1 info token)
- **Hint:** `1HR`, `1HG`, `1H1`, `1H2`, `1H3` - Tell the AI about their Red/Green cards or their 1s/2s/3s

## Model
The AI uses [nph4rd/Qwen3-1.7B-Tiny-Hanabi-XML-RL-12-2](https://huggingface.co/nph4rd/Qwen3-1.7B-Tiny-Hanabi-XML-RL-12-2), a Qwen3-1.7B model fine-tuned with reinforcement learning on this Tiny Hanabi environment.