Google postpones the release of its most powerful AI: the major flaw blocking everything internally

This is a slow start that is beginning to cost the Mountain View-based company dearly, enough to knock Alphabet’s share price down 4% on the stock market. Unveiled with great fanfare last May at the annual Google I/O conference, the most powerful version of its artificial intelligence model, Gemini 3.5 Pro, was expected to follow shortly thereafter, in June 2026. Its role was to support the streamlined version, named Gemini 3.5 Flash.Several months later, Google’s flagship remains conspicuously absent, and the reasons for this blockage have just been brought to light by a Bloomberg investigation.

Computer code, the Achilles heel of AI

If Google currently refuses to bring forward any official release date, content merely to assert, through its spokespersons interviewed by CNBC, that the model is still in the testing phase with certain partners, the internal situation seems much more tense. The American company’s engineers and researchers would be particularly disappointed with the capabilities of Gemini 3.5 Pro in programming and code generation. Despite a massive update to the training data carried out at the end of June, the results obtained in the laboratory remain stubbornly below the requirements set by management.

This technical shortcoming comes at the worst possible moment for the search giant. Computer code has indeed become the lifeblood and the main sales pitch for monetizing AI with developers and businesses. While Google is faltering, the competition is advancing at a breakneck pace. OpenAI has just launched its formidable GPT-5.6 Sol model, claiming superior technical efficiency on coding tasks, while Anthropic appeals to professionals with its powerful Fable 5. Even Meta joined the dance last week with Muse Spark 1.1, presented by its teams as the most robust open-source tool for autonomous development work.

Administrative bureaucracy weighing down innovation

Beyond mere technical obstacles, this delay would also shed light on chronic organizational problems within Alphabet. Developing the Gemini ecosystem now involves a multitude of teams and decision-makers, each seeking to integrate AI into their own division, whether YouTube, Maps, or Search. This multiplication of hierarchical approvals significantly slows the release of updates.

Some internal voices argue, moreover, that Gemini’s primary objective should not be fierce competition in programming aimed at experts, but rather the direct improvement of consumer-facing features, such as the famous AI Overviews integrated into search results. Until Google settles on its strategy and succeeds in bringing its model up to the level of its rivals, the company is asking users to be patient with the current 3.1 version.

👉🏻 Follow tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.