How to use Kimi K2 in Factory

Kimi K2 is a family of open-weight language models from Moonshot AI. Factory offers Kimi K2 natively, so you can run it inside a Droid without standing up your own inference or bringing a provider key. If you have been looking for where to use Kimi K2 for coding, this is how it works in Factory.

Kimi K2 sits in Factory's model list next to the Anthropic, OpenAI, Google, and xAI models. You select it the same way you select any other model, and Factory serves the inference and handles billing.

What Kimi K2 is

Kimi K2 is an open-weight model series, meaning its weights are published rather than served only through one vendor. Factory groups it under Droid Core, the set of open models it hosts and serves directly. Three variants are available today: Kimi K2.7 Code, Kimi K2.6, and Kimi K2.5. Kimi K2.7 Code is the coding-focused build in the family. All three are listed on the Available Models page with their IDs and settings.

Because Factory serves these models, Kimi K2 works across the surfaces where Droids run, with no GPU or serving stack to manage on your side.

Selecting Kimi K2 in a Droid

In an interactive session, run /model and choose the Kimi variant you want, or cycle models with Ctrl+N from the chat input. All interactive commands are documented in the Droid CLI reference.

For non-interactive runs, pass the model ID with -m:

droid exec -m kimi-k2.7-code --auto medium "refactor the date helpers and update the tests"

Each Kimi variant exposes Off and High (the default) reasoning. Use Tab in interactive mode or -r in droid exec to set the effort. The current levels per model are on the models page.

Cost and the credit multiplier

Factory bills usage with a per-model credit multiplier. Kimi K2.7 Code and K2.6 are listed at 0.4x, and K2.5 at 0.25x, which places the family among the cheaper options in the catalog. Confirm the live multiplier on the Available Models page before standardizing, since the numbers change.

The table below summarizes the Kimi K2 options as published.

When to reach for Kimi K2

Kimi K2.7 Code is worth trying on scoped coding tasks where you want an open model at a low multiplier. Keeping the option to switch models is the argument for a model-agnostic setup, and Factory lets you change models per task rather than committing to one provider. If you would rather have Factory choose the model for each step, Factory Router routes across the catalog automatically.

Kimi K2 in headless runs and Missions

The same model IDs work in CI and automation. A GitHub Actions step can run droid exec -m kimi-k2.7-code to keep a task on a low-multiplier model. In Missions, you can assign Kimi to worker agents with --worker-model kimi-k2.7-code while keeping a stronger model for validation, spending credits where reasoning matters most.

Further reading

• How to use GLM-5.2 in Factory
• How to use MiniMax in Factory
• Why your AI agents should be model-agnostic
• Available Models