Antim Labs

We trained a language model on games. Here's what we found antimlabs.com/blog/replay

We're launching early access to Gizmo, our automated sim creation tool. From text and/or image inputs, our agent generates SimReady assets and scenes from dimensioned primitives, with correct affordances and articulation.

We'll keep evaluating frontier models as they are released, and expanding the leaderboards to smaller VLMs with better latency. VGBench is live at antimlabs.com/vgbench.

The gap is clear: models show flashes of vision-driven competence then break on navigation loops, ambiguous controls, pathfinding, or basic drag-and-drop.

We are excited to launch VideoGameBench on Antim Labs, created by @a1zhang, Thomas L. Griffiths (@cocosci_lab), @karthik_r_n, and @OfirPress at Princeton.

more context: based on this and 3 other images, it created this sim (in isaac). tool's not perfect yet and we're looking for feedback to improve it

we’re opening this up to testers. it now supports both isaac and mujoco. if you’re interested in automating sim creation, reach out

we’ve been building an internal Isaac Sim tool that lets us prompt sims. swapped in gpt-5.4 and it one-shotted every asset in this datacenter scene. @openai cooked

@NotOlekslol @AnthropicAI 🫡

@AntimLabs Train it on Minecraft. @AnthropicAI still fucks up my builds

We trained a language model on games. Here's what we found antimlabs.com/blog/replay

Last day at #NeurIPS2025. If anyone wants to talk World Models, RL Spatio-Temporal Environments, Building General Intelligence using interactive 3D data, please hmu or the @AntimLabs team!

The Antim Labs team will be at #NeurIPS2025 from December 4th to 8th. If you’re exploring multi-agent RL, world models, or spatio-temporal reasoning, we’d love to connect. DM's open!