The highly anticipated new artificial intelligence model Gemini 3.5 Pro missed its planned release schedule due to significant technical problems encountered in code production. Google CEO Sundar Pichai promised at the developer conference in May that this tool would be released in June.

Although time has passed, the lack of an official update from the company has raised question marks about the future of the project. A new report revealed that the tech giant reset and updated core training information in late June to fix coding performance problems.

Institutional structure and internal competition slow down the process

The technical and operational setbacks experienced are largely associated with Google’s massive and modular corporate structure. Different groups in Google Cloud, DeepMind and Android departments are simultaneously working on developing their own AI programming tools.

This segmented approach increases internal competition, leading teams to compete aggressively for finite hardware resources. Google is working hard to combine these dispersed internal tools into a single umbrella platform called Antigravity.

The coding glitch comes at a time when Google’s dependence on automation was at its highest. Approximately 75 percent of new code deployed within the company is now generated by artificial intelligence and approved by engineers.

This rate, which was at 50 percent last autumn, clearly reveals the role of artificial intelligence in software processes. Many experts within the company argue that critical systems must be human-written to maintain stringent quality standards.

Google is struggling as competition heats up

While Google strives to perfect its flagship model, simpler initiatives in the artificial intelligence market are rapidly continuing their product development processes. Anthropic recently attracted attention by introducing its advanced Fable 5 architecture.

OpenAI has strengthened its position in the market by launching the GPT-5.6 Sol model, optimized for the coding and cyber defense domains. China-based Moonshot, on the other hand, joined the global competition with a huge open source model with 2.8 trillion parameters called Kimi K3.

Ability to code for truly pioneering AI models; It is considered the basic criterion of success because it requires logical reasoning, memory management and precise application. Google confirmed that it is continuing to test the Pro model with its corporate partners and the United States government.

It seems obvious that launching a faulty flagship model would damage the professional prestige of the brand. The real threat is considered to be institutional fragmentation between DeepMind and Google Cloud, rather than just an algorithm flaw.

Do you think this corporate structure of Google leaves the company behind its competitors in the artificial intelligence race?