An arena survival game where Jev or Laya controls every enemy
Sree's game describes each moment in one sentence and asks one typed question; the enemies act on the returned probabilities, with either model plugged in.
BUILDS
29 builds made with System One models.
Agents, routers, scorers and games — built on models that decide. Every entry links to the models it uses, so you can pull the same one and start from there.
19 builds
Laya plays Breakout at ~70 decisions a second on an RTX 5080
Ahmed had the 322M multilingual checkpoint control every paddle move, running locally on Windows.
Maze demo: 19 ms locally vs ~300 ms over the API
Dev Ed measured local Laya on a base M5 against Jev's hosted API on the same maze.
A Snake game steered by local Laya
AmirTaha feeds the fruit, head and wall positions to the model and moves in the direction it returns.
Tic-tac-toe against Laya in Unity
mgear runs the ONNX weights in Unity through onnxruntime-unity and plays full games against the model.
Jev and Laya race the Chrome dinosaur
Viraj Bhartiya ran Laya locally on MLX and Jev over its API on the same T-Rex track and physics.
Laya vs Jev in a Doom deathmatch
Hope set up a first-to-five 1v1: Jev's calls were sharper, while Laya decided about twice as often.
Offline Snake: Laya 43, Jev 1
With the network off, local Laya made 86.4 decisions a second from a 1 GB footprint while the API-bound Jev managed 3.2.
Local Laya vs cloud Jev at Tetris
atomic.chat pitted Laya on a 16 GB MacBook Air against Jev's API; the local model decided about 11× faster.
Lunar lander: Laya plus an MPC veto lands every time
Mansour Raad's experiment: Laya alone landed 0 of 90 attempts; with model-predictive control overruling a small share of moves, 90 of 90.
Pac-Man on Laya-MLX
soybelli connected Pac-Man to the model on Apple Silicon, with a live view of action probabilities, latency and tactical planning.
Laya-MLX ties a custom JAX build on a local game benchmark
Tom Siwik argues that framing the decision space and batching matter more than the model: over the network Jev wins, locally the two tie.
ScreenQuest: a fully local game agent on an M3 Max
Daniel Tremer's agent plays Hordes.io from screenshots: Apple Vision reads the screen, Qwen on MLX plans and Laya on Core ML makes each decision.
Thirty seconds of Snake: local Laya 46, cloud Jev 1
Simplifying AI ran the same game and questions on both models; Laya made 86.5 decisions a second against the API's 3.2.
A home-made Snake benchmark: Jev vs Laya-MLX vs Kev-4B
Tobias ran all three on a low-memory MacBook: Jev won on quality, the local models on speed and cost.
Tetris bench: Laya decides fast but loses to Jev
Tony Dinh ran laya-mlx at about 84 ms per decision against Jev in real-time Tetris; Jev won all three games.
World Summit Tournament: a fighting game with a Laya opponent
Daniel F's browser fighter has 21 characters, special moves, combos and replays, with the computer opponent run by Laya.
laya-mlx plays Snake at 60 decisions a second
The announcement of the MLX port: under 1 GB of RAM, choosing the snake's next move sixty times a second.
A day of RL training Laya on Doom
Shantanu Goel trained Laya for about 24 hours on FreeDoom's first level from text-only observations; zero-shot Jev still played far better.
Many of the first builds here were collected by madewithlaya.com. Every entry links to its creator’s own post, repository or site.