Models

SQLsaber supports many LLM models across different providers.

This guide covers model selection, configuration, and recommendations for the best SQL query generation.

Default Model

SQLsaber uses Claude Opus 4.6 by default, which provides excellent SQL generation capabilities with good speed and accuracy.

Recommended Models

While SQLsaber supports a lot of models, we recommend the following for best results:

• Any Claude Opus or Sonnet
• Gemini 3.1 Pro, 3.5 Flash or later
• Codex-5.3, 5.2, GPT-5.4, 5.5 or later

Tip

Consider using one of the recommended models for best results.

Model Management

List Available Models

saber models list

Set Your Default Model

Change the default model:

saber models set

This interactive command lets you:

1. Choose from configured providers
2. Select a specific model
3. Set it as your default
4. Configure thinking

Tip

This list is long - type to search and filter the list.

For scripts and coding agents, set the model and thinking level directly without fetching the model list or opening a prompt:

saber models set openai:gpt-5 --thinking-level medium

Subagent Overrides

You can override the model used by subagents like handoff and viz without changing the main model:

saber models set --agent handoff
saber models set --agent viz

# Direct, non-interactive override
saber models set openai:gpt-5 --agent handoff

If no override is set, the subagent uses the main model.

Reset to Default

Return to the original default (Claude Opus 4.6):

saber models reset
saber models reset --yes

Getting Help

Check your current model and available options:

saber models current
saber models --help

What’s Next?

After configuring your models:

1. Learn different ways to query your data
2. Manage conversation threads
3. Explore the command reference