How to Build the World's Fastest AI Game Generator with Qwen + Cerebras

How to Build the World's Fastest AI Game Generator with Qwen and Cerebras

In today's rapidly evolving AI landscape, creating interactive applications that generate content in real-time has become increasingly popular. What if you could build an AI-powered game generator that creates playable HTML5 games from simple text descriptions? That's exactly what we'll explore in this tutorial - building the World's Fastest nGame Generator using Alibaba's Qwen3-Coder model and Streamlit.

Introduction: What is This Project About?

Instantly turn text prompts into playable games with the world’s fastest AI game generator — powered by Cerebras Systems and the Qwen-235B model.

• Describe your game idea (e.g., “Flappy Bird with rockets”) and see it live in seconds

• Built on Qwen-235B with 256K token context (scalable to 1M) — perfect for large codebases

• Powered by Cerebras Systems with blazing 1,500–2,000 tokens/sec inference speed

• Generate and edit HTML5 games in real-time using a Streamlit-based chat interface

• No coding required — just prompt, play, and tweak instantly in your browser

<iframe src="https://drive.google.com/file/d/FILE_ID/preview" width="640" height="480" allow="autoplay"></iframe>

AI Integration: Cerebras + Qwen3-Coder

from langchain_openai import ChatOpenAI

def get_llm(api_key, model_name):
    return ChatOpenAI(
        model=model_name,
        openai_api_key=api_key,
        openai_api_base="https://api.cerebras.ai/v1"
    )

This setup allows us to connect to the Cerebras API for Qwen3-Coder access.

Architecture and Workflow

Step 1: Project Setup

mkdir qwen-game-generator
cd qwen-game-generator
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows
pip install streamlit langchain-openai

Step 2: Create Requirements File

streamlit
langchain-openai

Step 3: Main Application Code

import streamlit as st
from langchain_openai import ChatOpenAI
import os

st.set_page_config(page_title="🎮 World's Fastest Game Generator", layout="centered")

st.sidebar.image("https://github.com/pratik-gond/temp_files/blob/main/image-removebg-preview.png?raw=true", use_container_width=True)
st.sidebar.header("⚙️ Configuration")
model_name = st.sidebar.selectbox("Model", ["qwen-3-235b-a22b-instruct-2507"], index=0)

try:
    api_key = st.secrets["CEREBRAS_API_KEY"]
except KeyError:
    st.error("⚠️ Cerebras API Key not found in secrets.")
    api_key = None

st.sidebar.markdown("### About Qwen Model")
st.sidebar.markdown("Powered by Qwen model from Cerebras, delivering lightning-fast AI responses at **1500 tokens per second**.")

@st.cache_resource(show_spinner=False)
def get_llm(api_key, model_name):
    return ChatOpenAI(
        model=model_name,
        openai_api_key=api_key,
        openai_api_base="https://api.cerebras.ai/v1"
    )

if "messages" not in st.session_state:
    st.session_state.messages = []
if "game_code" not in st.session_state:
    st.session_state.game_code = ""

st.title("🎮 World's Fastest Game Generator")
st.markdown('*Built by [Build Fast with AI](https://buildfastwithai.com/genai-course) - Learn to build AI apps like this!*', unsafe_allow_html=True)

Step 4: Chat Interface Implementation

if api_key:
    for msg in st.session_state.messages:
        with st.chat_message(msg["role"]):
            st.markdown(msg["content"])

    user_input = st.chat_input("Describe your game or suggest an improvement...")

    if user_input:
        st.chat_message("user").markdown(user_input)
        st.session_state.messages.append({"role": "user", "content": user_input})

        is_game_creation = "<html" not in st.session_state.game_code.lower()

        if is_game_creation:
            full_prompt = f"""
You're an expert HTML5 game developer.

Generate a visually appealing and playable HTML5 game based on the following idea.
Requirements:
- Canvas-based game
- Retry button after losing
- Entire game in one standalone HTML file
- NO markdown (no triple backticks)

Game idea: {user_input}
"""
        else:
            full_prompt = f"""
Improve the following HTML5 game based on the user request.

Request: "{user_input}"

Only return a full, standalone HTML file (no explanations or markdown).

Game code:
{st.session_state.game_code}
"""

        with st.chat_message("assistant"):
            with st.spinner("🧠 Thinking... Generating your game..."):
                try:
                    llm = get_llm(api_key, model_name)
                    response = llm.invoke(full_prompt)
                    game_html = response.content.strip()

                    if "<html" not in game_html.lower():
                        st.error("❌ Invalid HTML received. Try rephrasing your request.")
                    else:
                        st.session_state.game_code = game_html
                        st.session_state.messages.append({
                            "role": "assistant",
                            "content": "✅ Game updated! See below 👇",
                        })
                        st.rerun()

                except Exception as e:
                    st.error(f"Cerebras API Error: {str(e)}")

Step 5: Game Display and Download

    if st.session_state.game_code:
        st.divider()
        st.subheader("🎮 Your Game")
        st.components.v1.html(st.session_state.game_code, height=600, scrolling=False)

        st.download_button(
            label="⬇️ Download Game HTML",
            data=st.session_state.game_code,
            file_name="ai_game.html",
            mime="text/html"
        )
else:
    st.error("Please add your Cerebras API Key to .streamlit/secrets.toml")

Step 6: Configuration Setup

# .streamlit/secrets.toml
CEREBRAS_API_KEY = "your-cerebras-api-key-here"

Advanced Features and Customizations

Adding New Models

model_name = st.sidebar.selectbox("Model", [
    "qwen-3-235b-a22b-instruct-2507",
    "gpt-4o-mini",
    "claude-3-5-sonnet"
], index=0)

Customizing Game Requirements

full_prompt = f"""
You're an expert HTML5 game developer.

Generate a visually appealing and playable HTML5 game based on the following idea.
Requirements:
- Canvas-based game
- Retry button after losing
- Sound effects (if possible)
- Mobile-responsive design
- High score tracking
- Entire game in one standalone HTML file
- NO markdown (no triple backticks)

Game idea: {user_input}
"""

Adding Analytics

if st.session_state.game_code:
    st.metric("Games Generated", len(st.session_state.messages) // 2)
    st.metric("Current Game Size", f"{len(st.session_state.game_code)} characters")

Error Handling and Validation

try:
    api_key = st.secrets["CEREBRAS_API_KEY"]
except KeyError:
    st.error("⚠️ Cerebras API Key not found in secrets...")

if "<html" not in game_html.lower():
    st.error("❌ Invalid HTML received. Try rephrasing your request.")

Running the Application

pip install -r requirements.txt
streamlit run app.py

Conclusion

The application bridges the gap between AI capabilities and practical game development. By leveraging Qwen3-Coder's code generation and Cerebras' fast inference, we created a tool that's both powerful and accessible.

Resources and Community

Join our community of 12,000+ AI enthusiasts and learn to build powerful AI applications! Whether you're a beginner or an experienced developer, our resources will help you understand and implement Generative AI in your projects.

• Website: www.buildfastwithai.com

• LinkedIn: linkedin.com/company/build-fast-with-ai/

• Instagram: instagram.com/buildfastwithai/

• Twitter: x.com/satvikps

• Telegram: t.me/BuildFastWithAI