Moonshot AI's Kimi launches game development package with K3.1 agents inside Unity, Unreal Engine and Blender
Moonshot AI has released a game development and creative work solution built on the Kimi K3.1 model series, Kimi Code, Kimi Work and plugins for Unity, Unreal Engine and Blender. Kimi says one Unity demonstration ran for 20 hours without human intervention.
Moonshot AI has released a game development and creative work solution for Kimi that runs K3.1 agents directly inside Unity, Unreal Engine and Blender. The package combines the Kimi K3.1 model series, the Kimi Code and Kimi Work agent products, and plugins for professional tools. Kimi says it is meant to help studios test gameplay ideas sooner, move engineering and art production forward, and feed player feedback from launch and live operations into the next development cycle.
How the agents work
According to Kimi, the model reads design documents, screenshots, code and 3D assets, then plans the work. Kimi Code handles implementation and changes inside a codebase. Kimi Work takes on multi-step production tasks that span several files. Through plugins and tool calls, Kimi operates on existing projects and returns editable files along with a record of what it did.
In Unity and Unreal Engine, once Kimi has the project files and necessary permissions, it first maps the code structure, scene hierarchy and module dependencies. It then makes changes through editor plugins and checks the result against reproduction steps, logs and runtime data. Code diffs and verification records are kept for engineers to review.
Unity China integration
For Unity, Kimi models work with Tuanjie Codely Agent, a game development agent from Unity China built into the Tuanjie Engine. Kimi interprets the developer's intent, writes code and analyzes screen recordings and test results. The Codely agent runs operations inside the engine through editor and runtime scripts. In one demonstration, Kimi says, the system worked for 20 hours without human intervention, recreating an existing game and producing a native project that can be developed further.
Showcase builds
The launch page lists several builds credited to Kimi K3.1 Pro:
- Blender: an indoor rainforest valley, rendered as a 17.5-second, 1920x1080, 60 fps flythrough.
- Blender (browser): a cathedral interior with about 3.3 million triangles and 8,368 animated components.
- Unreal Engine 5.8: a celestial palace scene with a colonnade of 36 bays on a 1,750-meter radius and a procedurally modeled dragon, totaling 36 million triangles rendered with Nanite and Lumen.
Kimi also says designers without coding experience can produce playable Godot prototypes in one to two days.
Beyond production
The solution also covers batch creation of ad creatives, in-game assistants connected through the Kimi API, sorting of community feedback with sources attached, localization drafts with back-translation checks, and first-pass screening of user-generated content. Kimi recommends starting with a single, well-bounded task and judging pilots on time saved, output quality and how much human revision is still needed. Pricing for the package is not yet disclosed.
What this means
The launch is a packaging of existing Kimi K3.1 models and agent products into specific engine workflows, not a new model. The practical emphasis is on auditability: Kimi says it retains code diffs and verification records for engineer review, and that it returns editable files rather than opaque output. The Unity path runs through Unity China's Tuanjie Codely Agent, which executes operations inside the engine, while Kimi supplies intent interpretation, code and analysis. Kimi's own guidance, to pilot one bounded task and measure time saved, quality and remaining human revision, gives studios a concrete way to evaluate the claims, including the 20-hour unattended run, against their own projects.
Related Articles
Fireworks claims up to 1,500 tokens/sec on DeepSeek V4.1 Flash in Ultra-fast inference preview
Fireworks announced a set of enterprise AI products at its first annual Forge conference on November 3, 2026. The lead item is an Ultra-fast inference preview that the company says reached up to 1,500 tokens per second on DeepSeek V4.1 Flash. It also announced a private preview of Fireworks Sovereign for customer-owned infrastructure.
Xcode 27.3 beta 1 adds GitHub Copilot to Apple's built-in coding assistant
Apple released Xcode 27.3 beta 1 on November 3, 2026, adding GitHub Copilot to the Xcode coding assistant. The beta also includes Swift 6.4, Device Hub improvements and UI testing support in Swift Testing. Xcode 27.3 is expected in December, according to 9to5Mac.
Xiaomi Humanoid Completes First 8-Hour Car Factory Shift; Xiaomi-Robotics-1 Enters Apartment Trial
Xiaomi says its humanoid robot worked an eight-hour shift at its car factory, covering the full daily production demand of the center console side-cover sorting station. The company also started a home-service trial of its Xiaomi-Robotics-1 embodied foundation model at a Xiaomi Youth Apartment.
Apple opens Siri AI to all third-party apps with 'app skills' in iOS 27.3 developer beta
Apple announced 'app skills' for Siri AI, allowing any app to create Siri AI integrations for the first time. The skills let apps teach Siri to coordinate complicated multi-step actions. The features arrive with iOS 27.3, macOS 27.3 and aligned releases, now in developer beta.
Comments
Loading...