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

A user argues Gemini 2.5 Pro writes more elegantly, while Gemini 4 Argon is full of trendy verbal tics, repeatedly saying "en vrai" in French.

Another Arena output from a new Gemini 4 Carbon checkpoint: a photorealistic 3D scene of thousands of nanobots assembling a futuristic spacecraft midair.

An Arabic post shows Gemini 4 Argon using three open-source datasets to create interactive 3D human anatomy visuals, with layer toggles during motion, muscle isolation and zooming inside a beating heart.

A leak claims Gemini 4 Flash is not Carbon, matches roughly Gemini 4 Argon performance while being much cheaper and faster, likely on all plans, with details next month.

A Korean user reports Gemini web and Antigravity track usage separately, and Gemini 4 Argon in Antigravity has its own weekly limit rather than sharing normal Gemini model quota.

The author says the latest Gemini 4 Carbon checkpoint in Arena outperformed kimi-k3.1 and qwen-4, praising Google's output since Gemini 4 Argon.

A user reports Gemini 4 Argon is down, with even a simple prompt taking a long time to respond, suggesting an outage or overload.

A video compares Gemini 4 Argon's second checkpoint against Opus 5.5 and Kimi K3.1 in Arena, with Argon coming out ahead.

A user posted a long video testing the first output of Gemini 4 Carbon's second checkpoint, calling its interfaces and tools cleaner than Argon and faster than v1.

An infographic lists 71 companies training their own models, noting Gemini 4 Argon holds #1 on LMArena's text leaderboard on a preliminary score, alongside xAI's rename and DeepSeek's OpenRouter share.

A user complains that a single tweet consumed $4 worth of Gemini 4 Argon tokens, hinting at high usage cost.

A developer rates coding feel across Sol, Opus 5.5, Fable, Gemini 3.8 Flash and Gemini 4 Argon, finding Argon's Japanese natural but Pro-plan quota too tight to judge coding.