Google’s flagship AI hits a snag: Gemini 3.5 Pro falls short of expectations

Google’s flagship language model, which everyone was expecting to see this summer, has hit unexpected technical roadblocks. According to multiple inside sources, Gemini 3.5 Pro is now delayed indefinitely — the AI simply can’t keep up with its rivals in one of the most high‑demand areas out there: coding.

Two months ago, the company rolled out lighter versions of the new generation — Gemini 3.5 Flash and Flash‑Lite — but the top‑tier Pro model never made it to users. Right now, the service interface still only offers the previous‑gen model, 3.1 Pro. Citing sources inside Google, outlets like NeoWin and Bloomberg are reporting that development is now months behind the original schedule, and the main culprit is weak performance on code‑writing and debugging tasks. The company’s engineers even gave the training dataset a major overhaul, hoping for a big leap in code generation quality. But internal testing produced the exact opposite effect: metrics didn’t just fail to improve — in some cases, they turned out worse than expected. And with OpenAI constantly updating its GPT lineup and Meta actively pushing its open‑source LLMs, falling behind like this is starting to hurt.

For free‑tier Gemini users, this news probably isn’t much cause for concern — their access to the Pro version is already capped anyway. But the delay hits paying subscribers the hardest. Those are exactly the users Google was counting on to show a return on its multi‑billion‑dollar AI infrastructure investments. When competitors are rolling out more capable alternatives, it’s only natural for users to wonder: why keep paying for a service that clearly lags in key use cases? And the longer this pause drags on, the higher the risk of developers and business clients jumping ship.

It’s worth noting that this isn’t the first time Google has found itself playing catch‑up. Back in the day, OpenAI’s ChatGPT got a massive head start while the corporate giant was busy shuffling names (PaLM, Bard, Gemini) and trying to figure out how to weave generative AI into search results. Back then, many chalked it up to big‑company caution. But now the stakes are much higher: enterprise clients are increasingly looking at objective code‑accuracy benchmarks, not flashy announcements. If they don’t get this fixed in the coming weeks, the gap could become nearly impossible to close.

For now, insiders aren’t predicting any release date, but one thing is clear: Google is going to need some serious retooling on Gemini 3.5 Pro to win back the trust of its professional audience and justify the massive spend on its AI ecosystem.

What do you think — does the Gemini 3.5 Pro delay make you want to take a closer look at what OpenAI or Meta have to offer, or are you willing to wait for the updated model from Google? Drop your take in the comments.