Infinity Arcade–Open-source local LLM showcase for generating games
Hi HN! We're Jeremy, Victoria, Krishna, and Daniel from the local LLM (Lemonade SDK) team at AMD and we're building Infinity Arcade (https://infinity-arcade.app). This project combines an open-source app and model to demonstrate what is possible for local LLM coding on the ~200 million everyday laptops out there with 16 GB of RAM. The progress has been surprisingly good, so I thought I'd show it here. The problem: while LLM coding is amazing with large cloud-based models, the results you get with today's open-source 7B-8B parameter models (the maximum size for an average consumer…
In plain words
Infinity Arcade is an open-source application and model that demonstrates local LLM coding capabilities on consumer laptops with 16GB of RAM. Developed by the AMD Lemonade SDK team, it uses smaller 7B-8B parameter models to generate working retro arcade games in Python. The project addresses the gap between cloud-based LLM coding performance and open-source model results, making local AI-powered game development practical for everyday computers.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hi HN! We're Jeremy, Victoria, Krishna, and Daniel from the local LLM (Lemonade SDK) team at AMD and we're building Infinity Arcade (https://infinity-arcade.app). This project combines an open-source app and model to demonstrate what is possible for local LLM coding on the ~200 million everyday laptops out there with 16 GB of RAM. The progress has been surprisingly good, so I thought I'd show it here. The problem: while LLM coding is amazing with large cloud-based models, the results you get with today's open-source 7B-8B parameter models (the maximum size for an average consumer laptop) are anything but inspiring. Even when we constrain ourselves to something "easy", like vibe coding retro arcade games in Python, the best sub-8B-parameter models can barely produce working copies of Snake and Pong. Our solution: first, we needed an app that would make it easy to drive local LLMs with exactly the right prompts and features for creating games. We built three minimal agents: Create, Remix, and Debug. This allows users to make new games, make changes to existing games, and allows the system to automatically fix its own bugs. Second, we needed to build our own model because none of the open-source offerings were cutting it. We assembled a dataset with >50k lines of high-quality Python game code, then used LORA SFT to create Playable1-GGUF (https://huggingface.co/playable/Playable1-GGUF): by far the world's highest performing 7B model at this task. It can handle 3x more game types and many fun riffs, like "space invaders with exploding bullets" and "breakout where the ball accelerates every time, I hit it." There's a short GIF with a demo of the outputs here: https://github.com/lemonade-sdk/assets/blob/main/arcade/arac... And one showing the full UI here: https://github.com/lemonade-sdk/assets/blob/main/arcade/infi... Ultimately, the goal is to deliver a reference design that inspires startups to build with local LLMs. Local has no ongoing cloud costs and no privacy/security concern with sharing user's data over the internet. The main challenges are the capabilities of smaller local models, as well as the work to integrate a software stack that is relatively less mature than OpenAI/Open Router APIs. We believe Infinity Arcade shows that both challenges can be overcome and the benefits of local can be realized. What you can do today: download the app from GitHub, install with 1 click, run Infinity Arcade and its LLM 100% free and local, check out the training data and process for the fine-tuned model, and build your own app using all of this as a reference. Links: GitHub: https://github.com/lemonade-sdk/infinity-arcade Model: https://huggingface.co/playable/Playable1-GGUF We’d love to hear your feedback! Are you convinced that the benefits of local outweigh the challenges? Do you have all the resources you need to get started? Let us know in the comments!
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, October 2025
the whole month →
- SA
I went down the rabbit hole on a side project and ended up building this: Strange Attractors(https://blog.shashanktomar.com/posts/strange-attractors). It’s built with three.js. Working on it reminded me of the little "maths for fun" exercises I used to do while learning programming in early days. Just trying things out, getting fascinated and geeky, and being surprised by the results. I spent way too much time on this, but it was extreme fun. My favorite part: someone pointed me to the Simone Attractor on Threads. It is a 2D attractor and I asked GPT to extrapolate it to…
AI · Oct 2025 · blog.shashanktomar.com
- ASAutism Simulator▲779
Hey all, I built this. It’s not trying to capture every autistic experience (that’d be impossible). It’s based on my own lived experience as well as that of friends on the spectrum. I'm trying to give people a feel for what masking, decision fatigue, and burnout can look like day-to-day. That’s hard to explain in words, but easier to show through choices and stats. I'm not trying to "define autism". I’ve gotten good feedback here about resilience, meds, and difficulty tuning. I’ll keep tweaking it. If even a few people walk away thinking, "ah, maybe that’s why my coworker struggles in those…
Life & fun · Oct 2025 · autism-simulator.vercel.app
