xAI IS COOKING: Grok 4.20 Just Dropped (2 Million Context Window)

xAI just dropped Grok 4.20 Beta and I put it through its paces inside OpenCode — here’s whether it can actually compete with Claude Opus, Google Antigravity, and Codex.

Grok 4.20 comes in at $2/M input and $6/M output, with web search at $5 per 5K queries — undercutting Google grounding significantly. It also claims a 2 million token context window, which may be a first for any model. Generation speed is genuinely impressive. The problem? It’s incredibly lazy. It keeps stopping mid-task and refuses to finish builds without hand-holding, which kills it for real agentic coding workflows. Compared to Gemini 3 Flash, Claude Opus, and Antigravity, it’s a meaningful step forward for Grok — but it’s not dethroning the top models yet. The Twitter/X live data angle via web search is worth watching though.

Try Harbor today:

https://harborseo.ai

Work with us:

https://incomestreamsurfer.com

Try Bright Data:

https://brightdata.com/?promo=incomestreamsurfers

Chapters:

0:00 First Impressions of Grok 4.20 Beta
0:10 Pricing: $2/$6 Per Million Tokens
0:23 Web Search vs Google Grounding Cost
0:45 2M Context Window & Gemini 3 Flash Compare
1:22 Harbor SEO (Sponsor)
2:33 Testing Grok 4.20 Inside OpenCode
3:00 The Laziness Problem
4:17 Generation Speed Is Insane
4:49 Final Build Walkthrough
5:14 Missing Features
5:34 Verdict: Better Than Claude?
6:03 Twitter/X Live Data Access
6:14 Wrap Up

#Grok4 #xAI #AICoding #VibeCoding #ClaudeVsGrok #OpenCode #LLM #AITools #DeveloperTools #AIModels

source