Google's answer to OpenAI's GPT-5 is coming soon: What we know

Google

Developers are reporting sightings of Gemini 3 inside Google’s AI Studio, fuelling speculation that the company’s next major AI model could debut as soon as this week.

Gemini 3 surfaces in early tests

Over the past few days, developers have noticed Gemini 3 references appearing in A/B tests across Google’s AI Studio. Screenshots and snippets of code shared on social media show new “flash-preview” builds being tested, while leaked benchmarks point to major improvements in coding and SVG generation tasks.

TestingCatalog and several developers on X claim the new model is already outperforming Claude 4.5, particularly in handling vector graphics and structured outputs. Meanwhile, Google Cloud’s official account has teased an October 9 livestream under the hashtag #GeminiAtWork, sparking widespread speculation that Gemini 3 could headline the event.

Why it matters

If accurate, Gemini 3 would represent Google’s biggest AI update of 2025, with expected upgrades in reasoning, coding accuracy, and creative generation. Analysts suggest it’s Google’s strongest answer yet to OpenAI’s GPT-5 and Anthropic’s Claude 4.5, both of which have set high benchmarks for AI performance.

A more capable Gemini could help Google regain ground in generative AI, especially if it’s deeply integrated into Workspace, Chrome, and Android. Developers, in particular, could benefit from faster, context-aware code generation and a rumoured “Agent Mode” that allows browser-level automation.

What’s next

Google has yet to confirm Gemini 3’s release, but all signs point to a major reveal at Wednesday’s #GeminiAtWork livestream. If the rollout begins, expect a gradual expansion through AI Studio and Workspace tools over the coming weeks.

For now, the tech world is watching closely — Gemini 3 could mark Google’s next leap in AI.