Google’s Gemini 3.2 Flash Goes Live in Secret – Code Generation So Powerful It Dwarfs Its Own Pro Model

Just before the I/O conference, Google silently launched Gemini 3.2 Flash, a new model that was first spotted by a Reddit user who noticed that the Gemini Canvas UI produced a dramatically different code style compared with Google AI Studio.

When users select the "Thinking+Canvas" mode, the backend silently routes requests to the new model, which appears in the Google Cloud Console as the entry gemini-3.2-flash-lite-live-preview. Many developers reported that this configuration reliably triggers Gemini 3.2 Flash.

The model’s coding ability is striking: where previous Flash models struggled to exceed 400‑500 lines, Gemini 3.2 Flash routinely generates over 1,000 lines and can produce up to 2,200 lines of code from a single prompt. Examples include interactive SVG graphics, a full Three.js project, a PS5‑style blueprint, and even a functional Windows 98 environment with a working browser, calculator, paint, Word, and Notepad.

In physical‑simulation 3D tests, Gemini 3.2 Flash generated high‑quality lighting, particle effects, and detailed PS5‑style SVGs with a single prompt, demonstrating a level of detail and interactivity previously unseen in Flash‑class models.

The breakthrough stems from aggressive model distillation combined with sparsification, allowing Google DeepMind to compress the essence of a large LLM into a lightweight version without the usual performance collapse.

According to internal benchmarks, Gemini 3.2 Flash reaches roughly 92 % of GPT‑5.5’s performance on core coding and reasoning tasks while reducing inference cost by 15‑20 times and keeping most query latencies under 200 ms.

Beyond raw coding, Gemini 3.2 Flash powers the Gemini App, which now integrates third‑party services such as GitHub, OpenStax, Spotify, WhatsApp, Canva, Instacart and OpenTable. Users can ask Gemini to design a wedding invitation in Canva, add ingredients to an Instacart cart, or reserve a table via OpenTable—all within a single conversational window.

Looking ahead, Google teases a suite of upcoming Gemini variants—Spark/Remy, Omni, Veo, 3.5 Flash, 3.5 Pro, Spark Robin, and Teamfood—each promising faster, cheaper, lower‑latency AI experiences. The imminent I/O 2026 event is positioned as Google’s chance to shift from “catching up” to leading the race toward artificial superintelligence (ASI).

Competitors are also advancing: OpenAI is preparing GPT‑5.6, Anthropic’s next model is on the horizon, and Claude Mythos remains a strong rival. Analysts note that while Google boasts unmatched infrastructure and product breadth, it still trails the top two models in raw capability, making the upcoming I/O a pivotal moment.