Models

Use the models you already trust

Ask five developers which model they reach for, and you’ll get five different answers. Claude for the complex refactor. A fast, low-cost model for boilerplate. A provider your company already pays for. A local LLM when code cannot leave the machine.

Zide lets you choose the model and access method that fit the task. Use included models, bring your own API key, link an eligible subscription, or run a compatible local model. Swap the model at will while your repository, files, Git state, and review process stay right where they are.

Different tasks, different models

A one-line fix and a 40-file migration can require different model capabilities.

Quick debugging calls for speed and low cost. Large-scale refactoring may require greater context capacity. Zide adds model selection directly into your workflow. Switch models mid-session without rebuilding your project context.

Zide's in-session model menu open over one Zide Assist session while a second session and a dev script sit in the tab strip beside it: the current model at the top, then a searchable list of Zide, Claude, DeepSeek, and GPT models showing when each was last used

Work across supported model providers

Zide supports leading frontier models, Zide Models, Bring Your Own Key (including OpenRouter), eligible model subscriptions, and compatible local models through Ollama and llama.cpp. Switch models mid-session without losing context.

  • ZideZideZide Models
  • AnthropicAnthropicClaude
  • OpenAIOpenAICodex
  • GeminiGoogleGemini
  • xAIxAIGrok
  • DeepSeekDeepSeekDeepSeek
  • Local LLMsRun models on your own hardware

Available models vary by provider and access method.

Zide's Crosscheck settings open over a conversation: Crosscheck switched on to review each new commit, with Claude Opus 5 chosen as the reviewer model while GPT-5.6 is the session's model
Crosscheck

Two models, one task

A second review can uncover issues the original model missed.

Crosscheck lets another model review the original model’s output before you commit. One model implements the solution and another reviews it. Use this additional review alongside your tests, Git diff, and existing engineering controls.

Access

Use the access method that fits your setup

Zide Models

Start immediately with built-in model access managed directly through Zide, including bundled usage credits.

Bring your own API key

Bring your own API key, including OpenRouter, to connect compatible models through your existing provider account.

Available on Zide Standard and above.

Use your own model subscription

Already paying for a provider subscription? Use an eligible model subscription inside Zide.

Run local models

Run compatible local models on your own hardware. Local model requests stay on your machine. Connect local runtimes like Ollama or Llama.cpp for completely offline execution.

Available on Zide Standard and above.

Privacy & data

Data handling varies by provider and access method

How data is processed depends on the model provider and the way the model is accessed. Zide Models, your own API keys, eligible model subscriptions, supported CLI agents, and local models may each be subject to different terms and configurations.

Zide Models

Zide-managed model access may use third-party model providers. See the Privacy Policy for information about how AI inputs, outputs, and related data may be processed.

Other access methods

When you use your own API key, an eligible model subscription, a CLI agent, or a local model, the applicable provider terms and configuration may differ.

Bring your own model, keep the workflow

Zide works with the models you already use, your own keys, or included credits. Everything else about how you work stays the same.