Wait Which Model?
← Back to Harnesses
xAINotable

Grok Build

Release date unpublished

xAI's coding agent: a terminal TUI, headless mode and Agent Client Protocol integration, plus an app-building mode in the Grok web and mobile apps.

Interface
CLI, WEB
Pricing
Subscription
Models
Any model
Best Terminal-Bench
37.58% (v4.0) · Grok 4.7
Open source
No
Licence
Proprietary
Terminal-Benchv4.0 · per harness + model pairing
Grok 4.737.58%
Grok 4.620.3%
Grok 4.512.42%
Strengths & weaknesses

Strengths

  • Interactive TUI, headless (`grok -p`) and ACP modes
  • Accepts any custom model via a base URL in ~/.grok/config.toml
  • Available on every Grok plan since 2026-08-19

Weaknesses

  • Grok 4.7 scores 37.6% on Terminal-Bench 4.0, well behind Claude Code and Codex
  • No open-source licence or repo published

Notable: appears on the official Terminal-Bench leaderboard (tbench.ai, Terminal-Bench 4.0 — the only version tbench.ai currently shows), best rank #21; only five distinct agents are listed there, so every one is top-10. Scores are TB 4.0 resolution rates (best per model across reasoning efforts), fetched 2026-10-14 — not comparable with older Terminal-Bench 2.x figures. releaseDate null: x.ai's 2026-08-19 post says Grok Build was introduced 'in July' as an Early Beta without giving a day. x.ai/build itself returned 403. 'web' covers the in-chat app builder on web/iOS/Android described in the same post.

Links