The next upgrade of Google's Gemini may not be released as soon as anticipated.

      Google played a crucial role in initiating the modern AI race, but maintaining its lead has proven to be more challenging than initially entering it. A recent report from Bloomberg indicates that the company is several months behind its internal timeline for launching Gemini 3.5 Pro, its upcoming primary AI model, as engineers work to enhance one of its major shortcomings: coding.

      The postponement is not merely about refining another chatbot; it underscores a larger issue for Google. The combination of extensive engineering teams, numerous product divisions, and increasingly stringent AI safety regulations is hindering the company’s responsiveness to competitors who seem to be progressing much more swiftly.

      As OpenAI, Anthropic, and Meta continue to release more advanced models, Google appears to be struggling to find the right balance between developing superior AI and maintaining the trust it has built with billions of users across its products.

      Coding remains the most significant hurdle for Gemini

      Bloomberg, referencing several current and former employees at Google, reports that the delay of Gemini 3.5 Pro is due to the company not achieving its coding performance enhancement goals. The report mentions that Google even updated the model’s training data late last month to enhance its coding abilities, but the outcomes reportedly fell short of internal expectations.

      This area is becoming increasingly crucial, as coding has become a key benchmark distinguishing today’s top AI models. OpenAI, Anthropic, and, more recently, Meta have heavily invested in AI systems focused on developers that can write, debug, and reason through complex software projects. According to the report, both OpenAI and Meta currently surpass Google’s available models in this domain.

      In defense of its progress, Google asserts that development is on track. A statement referenced by Bloomberg indicated that the company is testing Gemini 3.5 Pro, an enhanced version of its Flash model, along with other AI systems with partners while continuing dialogues with the U.S. government regarding testing standards and AI safety.

      The delay is also significant, as many anticipated Gemini 3.5 Pro would launch during this year's Google I/O event. Instead, the company concentrated on small increments of improvement for Gemini while competitors released new advanced models.

      Google’s substantial strength may also be a factor in its slowdown

      Unlike many AI startups, Google does not develop models in a vacuum. Each significant release of Gemini needs to function across various platforms like Search, YouTube, Maps, Android, Workspace, Cloud, and many more. While this scale grants Google tremendous advantages, including unparalleled access to real-world data, it also adds layers of internal coordination that can impede decision-making.

      Bloomberg's report details that current and former employees highlight competing priorities among teams such as DeepMind, Google Cloud, and Android, where overlapping AI coding initiatives complicate the maintenance of a unified strategy. Former employees also mentioned that internal disputes over AI-generated code and initial restrictions on the application of Gemini for software development limited experimentation during the early stages of the technology's rollout.

      Google maintains that those policies have evolved. The company asserts that approximately 75 percent of its production code is now AI-generated, and internal coding tools are being unified under a platform named Google Antigravity. It also notes that engineers are now expected to leverage AI for coding, although some still encounter computing capacity issues due to high internal demand for GPU resources.

      The report also indicates growing discontent within segments of Google’s AI organization, with some researchers reportedly moving to competitors like Anthropic. Meanwhile, customer opinions seem divided on Gemini 3.5 Flash. While companies such as Figma have lauded its combination of speed and quality, others, like the education platform Platzi, believe it occupies an uncomfortable middle ground, offering higher costs than previous Flash models without the reasoning capabilities of premium competitors.

      The overarching issue is that Google’s AI challenge has shifted from merely demonstrating its ability to build cutting-edge models—an aspect few doubt—to determining if a company of its size can deliver those models swiftly in an industry where competitors now gauge progress in weeks rather than months.