General Context Labs

[General Context Labs @generalcontext_]

Frontier models are solving Millenium Problems but still struggle to build complex 3D objects. How do we even measure what they’re getting wrong?

Introducing 3JSBench, the first open-source benchmark to evaluate LLMs on the ability to generate coherent @threejs assets.

[General Context Labs @generalcontext_]

@michaelzhao_ Thanks Michael!

[Michael Zhao @michaelzhao_]

a further step towards real world modeling from the awesome folks at @generalcontext_

[General Context Labs @generalcontext_]

@angelajiazhang Thanks Angela!

[angela zhang @angelajiazhang]

Exciting stuff from @generalcontext_ !! Big fan of this bench and everything else that will get us closer to real world deployments

[General Context Labs @generalcontext_]

@luckeyfaraday @threejs Thanks!

[General Context Labs @generalcontext_]

Frontier models are solving Millenium Problems but still struggle to build complex 3D objects. How do we even measure what they’re getting wrong?

Introducing 3JSBench, the first open-source benchmark to evaluate LLMs on the ability to generate coherent @threejs assets.

[General Context Labs @generalcontext_]

@Narfdemo2000 @threejs Nice!

[General Context Labs @generalcontext_]

@anmolbuildz @threejs Precisely

[General Context Labs @generalcontext_]

@jeffreygwang @threejs Astra is a wonderful model :)

[General Context Labs @generalcontext_]

@aphiticus1 @threejs Which models are the best with Blender?

[General Context Labs @generalcontext_]

@odpanda @threejs Astra is very strong in 3D

[General Context Labs @generalcontext_]

@sensho @threejs Astra is an insane model

[Scott @scottstts]

I like this bench, I think it aligns with my experience

- Astra is better at 3d modeling, whether it’s in threejs or blender, it’s especially good at very complex or subtle geometry

- Opus 5.5 is better at taste and art direction, it can produce things that just feels refined and polished with ease

In my workflow, I like to use opus 5.5 in general, and Astra for very specific modeling jobs

[General Context Labs @generalcontext_]

@scottstts This has been our experience exactly! Astra gets everything correct but sometimes feels sterile. Opus builds incorrectly every now and then but does feel injected with Taste

[Christopher Li @christopherrli]

@alexandr_wang wyt

[Frank Liu @frvnkliu]

Excited to share this one :)

[gary wu @gary_wu23]

It's incredible how many 3D AI game demos on X are just video dupes. The frontier models are still far from understanding and interacting in 3D

[General Context Labs @generalcontext_]

Frontier models are solving Millenium Problems but still struggle to build complex 3D objects. How do we even measure what they’re getting wrong?

Introducing 3JSBench, the first open-source benchmark to evaluate LLMs on the ability to generate coherent @threejs assets.

[General Context Labs @generalcontext_]

You can explore every task, inspect the verifiers, and interact with the generated objects yourself. Reach out if you’re interested in learning more or collaborating: [email protected]

https://t.co/VjiyZcJonf

[General Context Labs @generalcontext_]

Deterministic checks alone can’t tell you what looks good, so we collected 9500+ votes from 1000+ game creators who compare assets side by side. Users prefer Opus 5.5 slightly above Astra, despite Astra passing more of the correctness checks.