AI models"Did Google just dethrone OpenAI and Anthropic?" Fireship's 1.27M-view test drive of Gemini 4 Argon lands at 'Google tier'

Fireship, the developer channel with 4.31 million subscribers, posted "Did Google just dethrone OpenAI and Anthropic?" on Oct 20, and within a day it passed 1,268,000 views and 27,000 likes. The video pits Google's new Gemini 4 Argon against GPT-6 Astra and Claude Opus 5.5.

The video says Argon is a sparse mixture-of-experts model trained from scratch, not a fine-tune of older Gemini, and costs $2 per million input tokens and $10 per million output through Nov 30. But on Terminal Bench 4.0, a number Google didn't bold, Opus 5.5 wins 66 to 57, and Argon sits ninth in the human-ranked webdev Code Arena. "It's probably not going to be as good as Claude or GPT-6 at coding," Fireship concluded.

Hands-on results were mixed. An app promo video came out "mind-blowingly mid," and Claude's code-generated motion graphics for the same prompt looked better. Asked for a realistic 3D horse in Three.js, Argon added some anatomy nobody asked for, prompting the joke that he wouldn't feel safe using Gemini "until they fix the alignment."

Top comments included "TLDR: Googles model stay in the Google Tier" (2,600 likes) and "The only reason I pay for gemini is because it comes with 2tb of cloud storage" (2,700). Google said Argon processed nearly 25 trillion tokens in its first 72 hours, about 3x what Gemini 3 did.