LM Studio、Kimi K3 モデルを Bionic エンジンで実行可能に

We are excited to share that Kimi K3 is now available in LM Studio Bionic!

Our inference servers are US-based, with Zero Data Retention (ZDR) enabled by default.

Kimi K3 is Moonshot AI's most capable flagship model to date, and it's built for exactly the kind of long-horizon, agentic work that Bionic is made for. Kimi K3 is available as a cloud model in Bionic and runs under Zero Data Retention by default, like every cloud model we offer.

What is Kimi K3?

Kimi K3 is a 2.8 trillion-parameter model and the first open-source model to reach the 3T-parameter class. It performs competitively against many of the best proprietary models and represents a significant step forward in the open model ecosystem.

Kimi K3 features:

• 1M token context window so Bionic can hold large codebases, deep research, or large sets of documents in a single session

• Vision Support to understand images and screenshots, so you can tell Bionic exactly what you see

• Always-on reasoning with a choice between low, high, and max reasoning efforts, depending on the complexity of your task

Use Kimi K3 with the Bionic agent

1. Long-horizon coding in a Code project

Point a Code project at a local repository and choose Kimi K3 from the list of cloud models. Ask Bionic to search the codebase, read relevant files, and make changes. The model's long context and sustained reasoning make it well suited to complex coding tasks that span many files.

2. Frontend development

Kimi K3's native vision support lets you iterate with Bionic using images and screenshots. You can drag in an image, ask Bionic to make a change based on the image, and Bionic will reason about what it sees and refine its work based on visual feedback. Frontend development is one clear use case, but this also extends to non-coding tasks where you'd like to supplement your prompt with an image.

3. Deep research and knowledge work

A 1M-token context window means Kimi K3 can work with large amounts of information and files in one session. Create a Work project in Bionic and drag in a large folder of documents. You can ask the agent to pull out key findings, generate summary decks, and conduct deep research given the provided materials. The model's always-on reasoning helps it analyze and synthesize across files, while Bionic's automatic checkpointing lets you easily undo any unwanted changes along the way.

Select a reasoning effort

Here are some general guidelines for selecting a reasoning effort:

• Max: complex debugging, architecture design, and deep analysis or research

• High: everyday coding and document work

• Low: quick lookups, simple edits, and fast iterations

Get Started

Download LM Studio Bionic and create an LM Studio account to use cloud models. Cloud models in Bionic run on Zero Data Retention by default, so your data is not retained or used for training purposes.

We're excited to see what you build with it!