[Guide] After two full months of anticipation, Gemini 3.5 Pro is finally coming! Frontend code generated in one go, rumored to launch officially on July 17.
What might overturn Fable 5 could be Google’s secret weapon!
Today, the entire internet is flooded with leaks about Gemini 3.5 Pro, rumored to be officially released on July 17.
What really stands out is its front-end generation performance: one-time completion, pixel-perfect accuracy, and zero errors.
Most importantly, Gemini 3.5 Pro's performance has surpassed Fable 5.
Gemini 3.5 Pro's frontend capabilities surpass Fable 5
Google has held back its flagship product for two full months, and it’s finally about to be released.
At the I/O conference, the entire Gemini 3.5 series was unveiled at once:
However, only the lightweight Gemini 3.5 Flash was officially launched that day, while the premium Pro version remained in the background.
Pi Chai promised on stage, "Give us one more month."
June passed and Pro still didn’t arrive; now Google has changed its wording to "July," but it still hasn’t provided a specific calendar date.
Thus, during this period of official silence, leaks became the only source of information.
The Gemini 3.5 Pro assembled from these fragments presents a distinctly clear, yet notably specialized profile.
First, the conclusion—the strongest and most consistent signal from this leak:
Gemini 3.5 Pro showed a clear leap in capability in "frontend and visual code generation."
Based on feedback from multiple developers on X, the leaked version is visibly stronger in these aspects—
- Better design aesthetics: The generated interface is no longer "programmer-style"—colors, spacing, and hierarchy now reflect the precision of a professional designer.
- Cleaner UI: The frontend page structure is streamlined with less redundancy.
- SVG generation significantly enhanced: This was a frequently cited pain point—complex vector graphics can now be rendered correctly on the first try with minimal pixel loss;
- Frontend is more “highly completed”: generate a polished, fully formed page with a single prompt, rather than a half-finished draft.
During testing, developer QASIM-livelifewithai couldn't help but exclaim:
Gemini 3.5 Pro generated an SVG image identical to Logan Kilpatrick, head of Gemini!
A comparison image of Gemini 3.5 Pro and Fable 5 shows that the former has a slight edge in terms of mood and atmosphere.
For example, a single sentence can generate a UI.
Hand-drawn SVG mechanical solar system, also a prompt—Gemini 3.5 Pro accomplished it.
Gemini 3.5 Pro can also generate a floating steampunk island using Three.js.
In developer circles, a slightly playful term has even emerged to describe Gemini 3.5 Pro’s performance on the frontend—
Completely overpowering.
Hardcore reasoning is no match for Fable 5 or GPT-5.6
Dominating the frontend doesn't mean dominating everything.
In the anonymous tests on LM Arena, Gemini 3.5 Pro outperformed many competitors on these tasks.
However, in the most demanding agent tasks, warehouse-scale software engineering, and long-range tasks, 3.5 Pro still cannot match Fable 5 or keep up with GPT-5.6.
Fable 5 is a true powerhouse when it comes to heavy engineering tasks such as warehouse-scale, deep debugging, and high-ceiling architecture modifications, achieving highly competitive scores on benchmarks like SWE-Bench Pro.
GPT-5.6 also ranks among the top performers on the most challenging multi-step reasoning tasks.
Where does Gemini 3.5 Pro stand on this scale?
Even the leaker themselves admitted that, even with a new base, it still can't outperform Fable 5 or GPT-5.6 on the most challenging agents and long-chain tasks.
Earlier last month, a leaked benchmark even more bluntly ranked it after Fable 5 and GPT-5.6.
Its weaknesses are concentrated in reasoning, coding, and long-term task execution.
The base has been replaced; please retrain.
More intriguing than the frontend performance is why Gemini 3.5 Pro was delayed.
According to the leak, Google’s delay until July wasn’t spent on “tuning fine-tuning,” but on something more significant—
Performed a new pre-training, no longer using the older base model.
In other words, Pro is not simply a scaled-up version of Flash—it’s been rebuilt from the ground up.
In addition to the model itself, the leak also contains a hidden easter egg:
Google is developing a new image model called Nano Banana Pro based on the Gemini 3.5 Pro foundation, aiming to directly compete with OpenAI's GPT-Image 2.
If true, this means Google aims to fill gaps on both the text and image fronts simultaneously.
A single retraining session produced two product lines: one focused on frontend code generation and another on image generation.
Google has played its cards well—using the same foundation to fight on two fronts at once. Now, the entire internet is waiting for July 17.
And this is precisely what the 2026 large model arms race looks like in reality: everyone is racing to accelerate, but no one has managed to decisively overthrow their opponent.
Looking back over the past six months, the AI industry's buzz has had little to do with Google, as OpenAI and Anthropic have alternately stolen the spotlight.
But anyone who knows Google understands that it loves to hold back, accumulating everything before releasing it all at once.
This Gemini 3.5 Pro is likely another round of the same strategy—using a flagship model with a retrained foundation to make up for lost ground.
Whether Google will show up on July 17th, and whether it will truly be as leaked, the answer can only be revealed by Google itself.
But one thing is certain: in this battle of the gods, not a single player dares to pause for a breath.
Reference materials:
https://x.com/ggg78g89/status/2073783990646149488?s=20
https://x.com/HarshithLucky3/status/2073767413494284754?s=20
https://x.com/ggg78g89/status/2073852747049517231?s=20
This article is from the WeChat public account "New Intelligence Yuan," authored by ASI Revelation; edited by Peach.