Runchat Updates: December 18, 2025

📐 TRELLIS 2 and HUNYUAN 3

• These are both state of the art mesh generation models - can generate from 1536^3 voxel grids and with 4096px texture resolution
• Excellent at generating geometry with complex topologies and a lot of detail
• Relatively fast and low cost compared to previous state of the art (e.g. Sparc3d)

✨ GEMINI 3 FLASH

• Another state of the art model release from google
• Excellent at coding, tool use, general knowledge. Generally seems to be equivalent to Gemini 3 Pro but at much lower cost and higher inference speeds

🖼️ GPT IMAGE 1.5 and FLUX 2 PRO

• These are both pretty good editing models and excel at specific tasks

💳 USE GEMINI BILLING WITH ALL GOOGLE MODELS

• You can get $300 of free credit from google ai studio and use this in Runchat with Gemini, Imagen and Veo models
• We'll be posting a video tutorial on how to set this up sometime in the next few days, unfortunately it isn't a walk in the park but worth it for the credits.

We've also fixed a few bugs and made other minor improvements. Give us a shout if you want us to build something or add a default model!