Grok Build: user complaints meet a 17-point benchmark gap

xAI launched its terminal-native coding agent Grok Build on May 25, 2026 as an early beta. The model scores 70.8% on SWE-Bench Verified versus 87.6% for Claude Opus 4.7 Adaptive and 85% for Codex. Users report silent code changes and pricing up to $300 per month.

• SWE-Bench Verified: Grok Build 70.8%, Claude Opus 4.7 Adaptive 87.6%, Codex 85%
• SuperGrok Heavy costs $300/month, promo tier $99, while Claude Code is $20
• Up to eight parallel subagents in separate Git worktrees plus a Plan Mode
• API pricing: $1 per million input tokens and $2 per million output tokens