Intelligence, purpose-built for games.
We build lightweight LLMs designed to run locally in games.
Small and performant
Our models are typically 700-800 MB and run comfortably on CPU and most consumer devices. Latency is reliably under 500ms with >10 words/sec.
Finetuned for games
We finetune our models so they are much more likely to follow your game design constraints and stay in character compared to generic larger models.
Easy configuration
Add your game's custom requirements to the model context to control the model output and personalize it to your game.
Unity and Unreal plugins.
Easy integration with Unity and Unreal (we plan to support more engines, reach out if you need something custom).
Upcoming Models & Features
We have a lot of new models and features planned. Join our Discord or email us if you want a new model or feature — we'll put it on our roadmap.
Structured Outputs
Models that output structured values so it can dynamically affect game outcomes and gameplay.
Cross-platform support
MacOS, Android, iOS Support with platform-specific features.
Narrative Tool Integration
Integration with common narrative tools and frameworks like Articy, Inkle.
TTS model optimization
Optimize on-device TTS models so your NPCs can voice generated dialogue.
Start experimenting with our free models.
It's super easy to set up, we promise.