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

A developer says their PC is broken and Google Jules is reworking their latest build, then asks when Argon is coming.

The author uses Gemini 4 Pro to produce a Cappadocia-themed video, showing its scene and image generation ability.

A rumor says OpenAI will release Dots 2.0 and GPT-6.2 Sol next week with a $1,000 plan and no Europe launch, but GPT-6.2 Sol reportedly still loses to Gemini 4 Argon.

Antigravity has added Anthropic's new Opus and Sonnet 5.5 models, while commenters ask when Gemini 4 will arrive.

A post with an image exclaims that Gemini 4 Argon is still coming, hinting at new activity or clues.

A post lays out Google's full AI stack, listing Gemini 4 Argon as the frontier reasoning layer alongside Antigravity, Nano Banana, Omni and Spark.

A user generated an Anitkabir-themed video with Gemini 4 Pro, billed as a zero-shoot, fully AI-made clip.

A user asks whether anyone has actually used Gemini 4 Argon since it launched with a benchmark showing it beat Astra and Fable.

A leak claims Google has confirmed it will not release Gemini 3.5 Pro and has moved straight to Gemini 4 Argon, which is not yet public and reportedly beats GPT-6 Astra and Opus 5.5 on some benchmarks while trailing on others.

A user shares a screenshot claiming Gemini 4 Argon now beats Opus 4.8 on Agent Arena, addressing earlier doubters.

Using Blender MCP and Lichtfeld Studio, GPT-6.1 Astra rebuilt one of the new ChatGPT agent dots with fur detail that holds up from every angle, which the author contrasts with Gemini 4 Argon's strength in large scenes.

A developer who has spent about two months praising Gemini 3.8 Flash fears that once Antigravity offers Opus and Sonnet, complaints about tight limits will overshadow that reputation.