Google’s most powerful Gemini update has been pushed back once more, and this time the company reportedly threw out the nearly finished version to start over. If you use Gemini or were curious about it, that means the big upgrade you may have heard about is still not here.
The Gist
- The top version of Gemini, called 3.5 Pro, was expected around July 17 and slipped again.
- Google reportedly scrapped the almost-ready model because it was not good enough at writing code.
- The delay rattled investors, and Google’s parent company saw its share price drop about 4%.
Have ChatGPT Recap This Article
ChatGPTMeet Gemini, and the Update That Keeps Slipping
Gemini is Google’s AI assistant, its answer to ChatGPT. It lives in the Google app, in search, and across many Google products, and it comes in a few versions, from light and fast to large and powerful.
The version everyone was waiting for is the most powerful one, named 3.5 Pro. Google first hinted at a June arrival, then a July date, and now it has slipped yet again. If you are still choosing an everyday assistant, that is a good reason to keep your options open, and our list of the five free AI chats worth trying first is a calm place to start.
What makes this delay stand out is the reason behind it. Google did not just need more time to polish, it reportedly decided the model was not on track and chose to rebuild it from scratch.
How Google’s Do-Over Actually Works
Building a top AI model is a long process, and near the end you can still hit a wall. Google reportedly had a version almost ready, tested it, and found it stumbled on two hard jobs: drawing complex layered graphics and handling long chains of automated steps.
Rather than ship something weaker than its rivals, the company scrapped that version and went back to the drawing board. In late June it even changed the material used to teach the model better coding, and the results reportedly disappointed.
This is a normal risk in a fast race. While Google resets, rivals keep shipping, as we saw when a new ChatGPT model rolled out worldwide. A do-over can pay off, but only if the finished model actually beats what is already out.
Keep learning on AI Noobies:
- Kimi Is the Free Chinese AI That Rivals ChatGPT
- Netflix AI Is Now Behind Around 300 Shows
- GPT-5.6 Deleted Some Users’ Files by Mistake
What “Not Good Enough at Code” Really Means
One phrase drove this whole delay: the model was weak at code. Code is the set of instructions that runs apps and websites, and being good at it has become a headline test for how smart an AI really is.
Even if you never write a line of code, this matters to you. The same skill that lets an AI write clean instructions is what lets it follow your requests step by step, like booking, sorting, or filling things in on your behalf.
That is why the coding weakness spooked investors. You can already see where Google wants this to go in features like Google’s AI doing tasks inside your apps, and a stronger model is what would make those tasks reliable.
How to See This in the Tools You Already Use
For now, nothing on your phone changes. The Gemini you have keeps working, and the lighter, faster version is still there for quick questions.
If you want to feel the difference the missing model would make, try a harder task with your current assistant. Ask it to plan a full trip, or to turn a messy list into a clean table, and notice where it slows down or gets confused. Those rough edges are exactly what a new model aims to fix.
The healthy way to read this news is patience, not worry. A delayed model is a sign a company would rather wait than ship something weak, and knowing that makes you a calmer judge of the hype.
Stay tuned on AI Noobies.
About the Author
