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

Google Antigravity is offering researchers a six-month free Ultra tier with higher usage limits, Gemini 4 Argon access and a science plugin covering 30+ life-science databases; applications are reviewed weekly.

A poll-style post asks which model users would trust on a very complicated codebase: Gemini 4 Argon or Claude Opus 5.5.

A user reports Gemini Argon interrupting a long proof to muse "let us pause and appreciate Theorem 7," veering off-task.

A chart ranks models by score and per-task cost, placing Gemini 4 Argon at 53 for roughly $1.99 versus Opus 5.5 at 58 for $5.98.

The crypto newsletter's AI section notes Google launched its most powerful model, Gemini 4 Argon, alongside BTC, ETH and SOL market news.

The author argues Opus 5.5's real risk isn't GPT-6.1 or Gemini 4 Argon beating it, but cheap models like DeepSeek V4.1 Flash handling more flagship-level work, undermining the $200/month subscription.

EnvolDev compared Gemini 4 Argon, Sonnet 5.5, GPT-6 Astra and Opus 5.5, finding Argon costs about a quarter of Opus and runs roughly 7x faster with strong quality.

A video compares scene generation across Gemini 4 Argon, Claude Sonnet 5.5, GPT-6 Astra and Opus 5.5; Argon was fastest and cheapest, with praised sunset, waterfall and mist.

A developer says they call Gemini Argon, Grok 4.7 and DeepSeek V4 from inside Claude Code to build a stock strategy simulation, finishing in five days work that once took a team months.

Google is offering 10,000 scientists its highest Antigravity usage tier plus its most advanced Gemini 4 Argon model for free.

The author used Gemini 4 Argon in Antigravity to one-shot a roughly 56 KB HTML meditative animation.

Third-party API reseller ClawAPI offers Gemini 4 Argon at $2/$10 with 1M context and $0.10/M cache reads, at an intro price through Nov 30.