Models
progrok exposes the Grok models available to your xAI account through the local
OAuth bridge. GET /v1/models is authoritative for the active
runtime; this page lists representative uses, not a fixed catalog. Run
progrok models --detail before wiring a long-running automation.
Core text and coding models
| Model | Type | Best for | Notes |
|---|---|---|---|
grok-4.6 | General | Default chat, tool calls, search, vision, and most automation. | Good default for OpenAI-compatible clients pointed at the proxy. |
grok-build-0.1 | Coding | Fast coding, debugging, and agentic development loops. | Important for Grok Build-style workflows because it can use the same OAuth-activated surface. |
grok-4.20-0309-reasoning | Reasoning | Deeper reasoning when latency is less important. | Use explicit reasoning settings where supported. |
grok-4.20-0309-non-reasoning | Text | Lower-latency text generation. | Useful when you want Grok output without extended reasoning. |
grok-4.20-multi-agent-0309 | Research | Deep research, synthesis, and high-effort search workflows. | Pair with --reasoning high or --reasoning xhigh for expensive investigations. |
Voice models
| Model | Use | Stability |
|---|---|---|
grok-voice-latest | Follow the current Voice model. | Rolling alias; behavior can move. |
grok-voice-think-fast-2.0 | Realtime production sessions and the progrok live default. | Pinned identifier for reproducibility. |
Media models
| Model | Type | Best for | Notes |
|---|---|---|---|
grok-imagine-image | Image | Image generation and reference-based editing. | Use through progrok image for file handling and output directories. |
grok-imagine-image-quality | Image | Higher-quality still image output. | grok-imagine-image-pro is a retired slug that redirects here; prefer the quality model ID in new docs and scripts. |
grok-imagine-video | Video | T2V, I2V, reference-to-video, edit, and extend. | Base model for every confirmed video-input workflow. Use progrok video for polling, local files, and downloads. |
grok-imagine-video-1.5 | Video | Preview I2V generation. | Alias: grok-imagine-video-1.5-2026-05-30. Official model metadata lists text+image input; live OAuth smoke passes I2V but rejects prompt-only T2V and reference_images. |
Why Grok Build matters here
The point of progrok is not only “chat with Grok.” It is to make Grok usable by
developer tools. grok-build-0.1 represents the coding side of that
surface: an agent or coding CLI can target a local OpenAI-compatible endpoint,
send a placeholder key, and let progrok attach the same xAI OAuth session used
by related tools such as Hermes Agent and OpenClaw.
Live model inspection
progrok models --detail
progrok capabilities --jsonUse these commands before automation. The first is human-readable; the second is intended for scripts and agents that need stable metadata about commands, ports, models, direct endpoints, and auth requirements.