PRODUCT UPDATE / AI coding · Visual iteration
Inkling 2 Small launches with game and Blender demos
The November 5 release brings visual feedback to coding and 3D scene work. Tinker inference and fine-tuning are available, while weights remain closed and production serverless access has a waitlist.

Developers exploring browser-game prototypes and 3D scene ideas through a model that can review visual feedback.
Thinking Machines Lab released Inkling 2 Small on November 5, with image input and visual feedback for coding tasks such as game prototypes and 3D scenes. It is available for inference and fine-tuning through Tinker, with closed weights at launch. The company plans an Apache 2.0 weight release the following week, after external safety testing and red-teaming.
A city simulator and a Blender scene
The release announcement shows a Three.js city simulator built with the model. Thinking Machines says its buildings, textures and cars are generated procedurally, without premade assets. Players can expand from a single road while managing power, budgets and emergencies.
A second example uses Blender to create a paper-like valley with origami cranes. The announcement describes revising the scene's water and chimney. Both examples illustrate a workflow in which the model can inspect visual output and make further changes, which could help developers iterate on a browser-game concept or explore a scene's appearance.
These are vendor demonstrations. A game project would still need its own checks for gameplay, performance and asset preparation. Inkling 2 Small accepts text and images and produces text; this version does not support audio.
Tinker access and token prices
The model can be tried in Tinker Playground and used for sampling or fine-tuning on Tinker. Downloadable checkpoints are planned to become available when the model's weights are released. The current pricing guide separately labels Serverless Inference as beta, with a waitlist for production use.
At the time of this release, the guide lists $0.30 per million input tokens, $0.03 for cached input and $1.00 per million output tokens for Inkling 2 Small. Training is billed separately: $1.50 per million tokens for the 64K-context model and $3.00 for the 256K-context variant. Those are token rates; the total cost of an iterative coding session depends on its usage.
The prototyping and coding category covers tools for building an initial playable version. For further scene work, the 3D assets and animation category provides related production options.
Sources (3)
Based on official announcements and public documentation. No sponsorship or affiliate links.




