Gemini 4 Argon Just Dropped: EVERYTHING You Need to Know (3 Use Cases Built)

Summary Christian Peverelli evaluates Google’s newly released Gemini 4 Argon model and its agent environment, Google Antigravity, testing them head-to-head against Anthropic’s Claude Opus 5.5 using Claude Code. He examines official benchmarks, pricing, and independent evaluations before conducting a three-round test: building a landing page from a scraped design system, automating a video hook edit, and creating a playable 3D browser racing game from a desk photo.

What is shown

• [00:20] Announcement posts and model specs for Gemini 4 Argon from Google CEO Sundar Pichai and Google DeepMind, including tier availability in Google AI Pro ($19.99/mo) and Google AI Ultra ($99.99/mo).
• [01:05] Gemini interface showing the switch from “Chat” mode to “Task” mode and integrations with external tools.
• [01:20] The Google Antigravity workspace interface running agentic workflows alongside the terminal and file viewer.
• [01:34] Google’s official blog post and benchmark scorecards comparing Gemini 4 Argon against GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5.
• [03:21] Task 1 (Landing Page): Peverelli tests generating a sales page replicating his site’s branding. Antigravity’s result [03:35] is contrasted with Claude Code’s result [04:35] (scored 6/10 for Antigravity vs. 8/10 for Claude Code).
• [05:32] Sponsored segment covering Hostinger’s hosting, domain, and agent deployment tools.
• [07:02] Task 2 (Video Hook Editing): Edit #1 produced by Claude Code [07:02] vs. Edit #2 generated by Antigravity [07:44] with motion graphics, captions, and b-roll (scored 6/10 for Claude Code vs. 8.5/10 for Antigravity).
• [08:47] Task 3 (3D Desk Racing Game): Antigravity’s playable 3D car game “Desk Grand Prix” [08:47] navigating a desk surface with microphone and cable-hole obstacles, compared to Claude Code’s “Desk GP” [09:23] recreating the surrounding room and furniture (scored 8/10 for Antigravity vs. 9/10 for Claude Code).
• [10:41] Comparative summary tables displaying build execution times and API cost estimates across both platforms.

Claims & numbers

• The presenter notes Gemini 4 Argon launched with access via the API, the Gemini app, and Google Antigravity (available to Ultra subscribers on the $99.99/mo tier, while Pro costs $19.99/mo).
• The presenter cites Google’s release materials claiming Gemini 4 digested a 500-page chemistry patent in an afternoon and translated over 100 pages into 17 languages in under two days.
• On benchmark performance, the presenter states Gemini 4 won 13 of 19 benchmarks listed by Google, including Vals Finance Agent v2 (65.4% vs. Claude Opus 5.5’s 58.6% and GPT-6 Astra’s 53.5%) and Harvey’s Legal Agent Benchmark (19.6% vs. Claude Opus 5.5’s 3.8%), but lost on Terminal-Bench 4.0 (57.4% vs. Claude Opus 5.5’s 66.4%).
• The presenter quotes the Artificial Analysis Intelligence Index placing Gemini 4 Argon at 53 (tied with ChatGPT at 53, behind Claude Opus 5.5 at 58), while rating task automation at 78% for Gemini 4, 70% for Claude Opus 5.5, and 68% for ChatGPT.
• Build times reported by the presenter:
  • Landing page: Claude Code (12 min) vs. Antigravity (approx. 20 min).
  • Video edit: Claude Code (22 min) vs. Antigravity (34 min).
  • Desk racing game: Claude Code (46 min) vs. Antigravity (55 min).
• Estimated API costs reported for the three builds combined:
  • Claude Opus 5.5: $17.00 total ($4.00 landing page, $6.80 video edit, $5.70 game).
  • Gemini 4 Argon introductory pricing: $14.00 total ($3.40 landing page, $5.90 video edit, $4.80 game), rising to an estimated $28.00 total after planned price doubling on December 1.

Notable quotes

• [00:03] “Holy smokes, they are back. And I’ve been glued to my screen ever since, building everything I could think of.”
• [05:22] “So, Claude Code 1, Antigravity 0.”
• [10:30] “Claude Code 2, Antigravity 1. We have a winner my friends.”

Assessment This is an independent hands-on product review and side-by-side benchmark test evaluating Gemini 4 Argon and Google Antigravity against Claude Code running Claude Opus 5.5. The tests showcase actual browser-rendered code, video edit clips, and playable 3D WebGL games built directly during the review, though the presenter acknowledges running on top-tier paid subscriptions ($100/mo Google AI Ultra and $200/mo Claude Max) rather than base tiers.

Described by gemini-3.8-flash on 2026-10-14 from the video’s audio and frames.