Gemini 4 Pro / Flash: benchmarks, hands-on tests and leaks

KrishnaBuildsAI says they are currently using Gemini 4 Argon and asks for one good reason to switch to Claude Opus 5.5.

A post claims Google's Gemini 4 Carbon is an upgraded Argon with dominant visual abilities in video analysis and website testing, boosted by YouTube data.

A user reports Gemini 4 Argon used ~4M tokens in Antigravity without finishing a coding task, while Claude Opus 5.5 used ~240K tokens and passed all 90 tests.

After two weeks and 5.8B tokens on Gemini 4 Argon, the reviewer calls it the best long-horizon builder, running 15–30 hours without drifting, yet self-scoring 9 versus 6–7 from an independent reviewer.

A roundup post claims Google's Gemini 4 Carbon, an Argon upgrade, keeps a vision edge in video analysis and website testing with help from YouTube data, alongside rumored Anthropic and OpenAI Terra models.

A recommended article for anyone using Gemini 4 Argon documents not only how a Kindle book was completed but also the failure that stalled at seventy percent.

A post jokes that while developers feared AI replacing them, video editors seeing Gemini Argon said "vendor, please take your PC back," with video attached.

A user claims Gemini 4 Argon outperforms GPT-6 Astra in real-world use cases, without giving specific test details.

A post claims Gemini 4 Argon can detect security vulnerabilities at a lower price point than competing models, with results shown in an attached image.

A user calls Gemini 4 Argon promising but its usage limits a joke, asking whether Argon is worth upgrading to while Claude and GPT remain competitive for coding.

ValsAI's updated index puts Sonnet 5.5, Opus 5.5 and Gemini 4 Argon in the top three, with DeepSeek-V4.1-Flash-1017 the leading open-weights model at #8.

The author argues Claude Opus 5.5's danger isn't being surpassed by GPT-6.1 or Gemini 4 Argon, but cheap models like DeepSeek V4.1 handling more flagship-level work, questioning the $200/month subscription.