If Gemini 4's benchmarks are real, Google just pulled off the biggest comeback in AI history. People are building crazy stuff with the rumored Gemini 4 Pro checkpoint in Arena 10 Wild Examples

Why is Spider-Man’s head on backwards in Gemini 4? 🫪

Is it even real? They did this the last time as well, getting credulous people like you to upsell a POS Won’t be fooled again.

Gemini 4 vs Astra on the same prompt.

Gemini 4 Pro crushes GPT-6 Astra.

Gemini 4 Pro beats Opus 5.5

Compared Opus 5 vs GPT Astra vs Gemini 5 in Arena vs Grok 4.6

Robotic Humming bird

GPT-6 Sol High vs Gemini 4

Same prompt, but Gemini 4 was better

Gemini 4 checkpoint is absolutely a beast

Vintage steam starts flying with its sails

P.S. I dive deeper into AI tools—and where they’re headed—in my newsletter, AI Toast. Join 35,000+ readers in the AI Toast community for free:

Bonus: another update you shouldn't miss Qwen3.8-Flash is free inside Qoder right now 🤯👇

Co? 🤣 Kto to zrobił? Opus 5.5 zrobił tego Spidermana a Astra jest niewiele gorsza.

is this entire thread satire

Biggest comeback, bold call. Still riding a rumored checkpoint, though.

Fun to watch a big model comeback, but I’m mostly curious how stable the rumored checkpoints are in real workflows. If the examples hold up, it could genuinely change what small teams can prototype.

如果属实，Gemini 4 的基准测试狂飙就是谷歌最惊艳的 AI 逆袭——彻底改写游戏规则。

lol

This is because they don’t have a vm for the models on arena to build real 3d games yeah? Either way great svg work