Models

Use the models you already trust

Choose the model that fits the task without moving the rest of your development workflow.

Zide Assist works across supported model options while your repository, files, terminal, Git state, pull requests, and CI stay in the same workspace.

Screenshot

The actual Zide model picker inside the workspace, with repository/task context visible so this reads as model choice inside real development work, not a model catalog.

Recapture after the desktop rebrand · 16:10

Different tasks can call for different models

A debugging task, refactor, migration, or code review may not benefit from the same model every time.

Zide keeps model choice inside the development workflow. Change the model for the task without rebuilding the project context around it.

Pick the model for the work. Keep the workflow in Zide.

Video loop

A simple model-selection moment beside the same active repository or task.

Recapture after the desktop rebrand · 4:3
Supported options

Work across supported model providers

  • Claude
  • Codex
  • OpenAI
  • Gemini
  • DeepSeek
  • Local models where supported
  • Additional supported models as available

Exact model availability can vary by provider, access method, plan, and current product support.

Screenshot

The live Zide model picker as the primary visual, with provider names in a restrained row or grid beneath it.

Recapture after the desktop rebrand · 16:9
Access

Use the access method that fits your setup

01

Zide Models

Use supported model access provided through Zide, including included AI usage where available.

02

Bring your own API key

Use your own API keys where the provider and current Zide integration support it.

03

Use your own model subscription

Connect supported model subscriptions where available.

04

Run local models

Use local models where configured and supported by your setup.

Privacy & data

Built for model choice without making privacy vague

For Zide Models:

Data handling may vary depending on the model provider and how the model is accessed. See our Privacy Policy for more information about how AI features and third-party providers process data.

This statement applies specifically to Zide Models. When developers use their own API keys, model subscriptions, or local models, the applicable provider terms and configuration can differ.

Video loop

Zide Assist with selected context, a reviewable diff, and sandbox controls in one sequence.

Recapture after the desktop rebrand · 4:3

Give the model context. Keep the decision with the developer.

Zide Assist works inside the development workflow rather than in a separate chat surface.

  • Add context from tabs, issues, files, diffs, terminals, or list items.
  • Review proposed changes as diffs.
  • Use sandbox controls where tighter limits are needed.
  • Keep Git, tests, pull requests, and CI in the normal engineering loop.

AI can assist with the work. The developer still reviews the result and decides what moves forward.

FAQ

Questions about models

Can I choose which model to use?

Yes. Zide is designed around supported model choice so developers can select the model that fits the task.

Can I bring my own API key?

Yes, where the provider and current Zide integration support it.

Can I use my own model subscription?

Yes, where supported by the current provider and Zide configuration.

Can I run local models?

Yes, where local models are configured and supported.

What happens to prompts and responses when I use Zide Models?

Data handling may vary depending on the model provider and how the model is accessed. See our Privacy Policy for more information about how AI features and third-party providers process data.

Do the same data terms apply when I use my own key or a local model?

Not necessarily. Data handling can depend on the provider, access method, and local configuration.

Model access varies by plan and configuration.

Compare Plans

Choose the model. Keep the workflow.

Open your repository, choose a supported model for the task, and keep the code, terminal, Git state, review, and delivery workflow in one place.