xAI
Lakehousecat supports xAI as a provider for Chat and Speech-to-Text (STT), using the Grok model family. This page describes how to configure xAI as a Provider Model.
Prerequisites
- Access to Lakehousecat as an Administrator
- An xAI API key (from the xAI console)
Configuration
Navigate to Workspace → Models → Provider Models and click the + icon.
Step 1: Select Provider Type
From the Provider Type dropdown, select xAI.
Step 2: Select Model Type
Select Chat or STT (Speech-to-Text), depending on what this configuration is for.
Step 3: Enter Configuration Details
| Field | Description |
|---|---|
| Configuration Name | A unique name for this configuration (e.g. xai-grok-chat). |
| API Key (required) | Your xAI API key. Used for authentication. |
| Model ID (required) | The Grok model identifier, entered as free text (e.g. grok-4.5 for chat, grok-stt for Speech-to-Text). |
| API Base URL (optional) | Override the default xAI endpoint. Leave empty to use https://api.x.ai/v1. |
| STT Language (optional, STT only) | Language hint for transcription (e.g. en, de). Leave empty for auto-detection. |
xAI model IDs are entered manually rather than chosen from a fixed dropdown — the same as Azure and AWS. This keeps your deployment from being locked to the models known at release time: when xAI publishes a new Grok model, you can use it immediately by entering its ID from the xAI model documentation, without upgrading Lakehousecat.
Tested Configurations
Stand: 2026-07-21. This reflects what Lakehousecat has actually run and observed, not a benchmark or performance ranking — see There is no single "right" provider for why we don't publish comparative scores.
| Model ID | Type | Chart Generation | Notes |
|---|---|---|---|
grok-4.20-fast (non-reasoning) | Chat | ✅ Recommended starting point (3/3 charts, ~24s/chart, ~$0.04/chart) | Non-reasoning variant — roughly 2.6× faster than the reasoning models tested, no quality difference observed for this structured task |
grok-4.5 | Chat | ✅ Validated (1/1 charts, reasoning model) | Flagship reasoning model — works, but slower and costs more than the fast variant for the same chart output |
grok-4.3 | Chat | ✅ Validated (1/1 charts) | Previous flagship generation |
grok-build-0.1 | Chat | ✅ Validated (1/1 charts) | Build-focused model |
grok-stt | STT | — | Speech-to-Text — validated end-to-end |
Step 4: Save
Click Save. After saving, you can optionally set this configuration as the default UI model or default Backend model (Chat type only).
Best Practices
- Use a descriptive configuration name that includes the model name and type.
- Keep your API key updated when rotating credentials.
- For chart generation,
grok-4.20-fastis the recommended starting point — it is the lightest of the tested models and produced results comparable to the heavier reasoning variants in testing, at a fraction of the time and cost. See Choosing a Model for Chart Generation.
Speech-to-Text (Grok STT)
To use xAI for voice input, create a Provider Model with Model Type = STT and enter the STT model ID (e.g. grok-stt). Optionally set an STT Language hint. xAI STT uses the same API key as Chat — no extra setup is required. Once saved, set it as the default Speech-to-Text model so that voice prompts are transcribed via xAI.
Transcription uses xAI's batch endpoint — audio is uploaded and transcribed in one request (no live streaming). This is the same apply-then-transcribe flow used for voice prompt input across Lakehousecat.
Troubleshooting
- Validation Errors: Verify the API Key and Model ID for accuracy.
- Model not found: Confirm the model ID against the current xAI model documentation — model names change as xAI updates its catalog.
- Custom endpoint: If you set an API Base URL, ensure it points to an OpenAI-compatible xAI endpoint.