Grok
Quick Facts
- Vendor
- xAI (Palo Alto)
- Released
- Grok-1 (2023, weights released March 2024); Grok 4 (2025)
- Current line
- Grok 4 · Grok Code Fast · Grok 3
- License
- Current flagship: proprietary API. Grok-1 weights: Apache 2.0.
- Hosting
- xAI API, X Premium integration
- Context window
- 128K–256K tokens
- Modalities
- Text, image; real-time X search as a first-class tool
- Training
- Colossus supercomputer (Memphis, TN)
Summary
Grok is xAI's LLM family, launched in late 2023 and positioned as an alternative to the incumbent frontier labs. The defining differentiator is first-party real-time search over X — the model can retrieve posts, threads, and trending context directly, without a separate retrieval pipeline. For workloads where current events and social signal matter, this is unique among frontier models.
The Grok Code Fast variant specifically targets agentic coding: lower latency and higher throughput than the flagship, with a pricing structure designed for high-volume IDE and agent usage. xAI has also released weights for older Grok versions (Grok-1 under Apache 2.0 in March 2024), though current flagship models remain API-only.
Model Lineup
- Grok 4 — current flagship. Long context, reasoning-capable, strong on benchmark leaderboards.
- Grok Code Fast — code-specialized, latency-optimized. Targeted at IDE and coding agent integrations.
- Grok 3 — prior flagship. Still available for stable pipelines.
- Grok-1 (314B MoE) — open-weights release from March 2024 under Apache 2.0. Research and self-hosted use.
Where Grok Fits
Grok is the default choice when real-time X data is a first-class requirement — social monitoring, trend analysis, news-reactive agents. Grok Code Fast is worth evaluating for high-volume coding agent workloads where per-token economics dominate. For teams already in the X / X Premium ecosystem, the integration path is shorter than stitching together Claude or GPT with a third-party X retrieval service.
Tradeoffs
- Mostly closed. Current flagship models are API-only. Only the older Grok-1 weights are open. Most regulated deployments will need to look elsewhere.
- Ecosystem maturity. Third-party tool integrations, SDK coverage, and community tooling lag the Anthropic / OpenAI ecosystems.
- Platform coupling. The X integration is a feature and a liability — workloads that depend on it are coupled to X's availability and policies.
- Brand / product risk. xAI ships quickly and visibly. Pin snapshots and monitor deprecation schedules carefully.
Deployment Notes
Within the Claw ecosystem, Grok is a specialist provider — integrated for workflows that need real-time X retrieval, or when Grok Code Fast pricing beats alternatives on a specific high-volume agent loop. It sits in the provider arbitrage layer alongside Anthropic, OpenAI, and open-weights endpoints. We don't recommend it as a primary provider for regulated workloads given the API-only posture of current flagship models.