Google delays Gemini 3.5 Pro launch by months as coding skills fall short

Alphabet's Google is months behind its internal schedule in delivering Gemini 3.5 Pro, the company's most powerful flagship AI model. According to people familiar with the matter, Google has been investing time to improve the model's capabilities, particularly in coding — a task at which rival models from OpenAI and Meta have recently demonstrated superiority. In a bid to narrow this gap, Google updated the training data for Gemini late last month, but the results proved disappointing.

The delay has become a source of friction within Google's AI organization. Ten current and former employees told Bloomberg that engineers, AI researchers, and managers are concerned the company risks losing competitive advantage as Anthropic and OpenAI produce models that exceed Gemini's current capabilities. The tension reflects a genuine capability gap: both OpenAI and Meta have recently released new models that outpace Google's current offerings specifically in AI for writing code.

Google's organizational structure contributes to the scheduling challenge. The company operates with multiple layers of stakeholders preparing models for release while simultaneously working to integrate AI across a vast product portfolio — search, maps, YouTube, and others. This complexity can cause delays that more focused competitors may avoid. One former employee described the coordination effort as "trying to boil an ocean." When mandates shift or efforts duplicate across departments, maintaining a cohesive release strategy becomes harder, and individual product teams struggle to secure the resources they need to gain traction in the market.

Publicly, Google has downplayed the setback. A company spokesperson stated: "We're shipping quickly across a wide range of models while keeping them highly cost-effective for customers." The company is currently testing Gemini 3.5 Pro and an upgraded Flash model with partners. Google is also in talks with the US government, which has been increasingly monitoring advanced AI models, about Gemini's capabilities and the industry safety standards that should apply. This regulatory engagement mirrors constraints faced by competitors: earlier this year, Anthropic temporarily pulled its latest models after internal testing flagged dangerous cybersecurity capabilities, and OpenAI has voluntarily limited and staggered its newest model releases under pressure from national security concerns and the Trump administration.

Google's months-long delay on Gemini 3.5 Pro signals a fundamental tension between the company's scale and its ability to move fast in competitive AI. The body reveals that Google has been attempting to close a specific technical gap — coding capabilities — that rivals Anthropic and OpenAI have already solved in their latest releases. This is not a matter of basic functionality but of competitive parity: the company's internal team recognizes that falling behind on a high-value task (code generation) risks market position against focused competitors.

The article attributes the delay partly to Google's organizational structure. Unlike OpenAI or Anthropic, which ship individual models under a clear decision-making hierarchy, Google must coordinate across "multiple layers of stakeholders" preparing the same technology for integration into search, maps, YouTube, and other consumer products. The body quotes an ex-employee: "encouraging leadership of every department to move in the same direction is like trying to boil an ocean." This is not a technical limitation but an alignment problem — when mandates shift or efforts duplicate across departments, maintaining a cohesive release strategy becomes nearly impossible.

The regulatory context adds another layer. The body notes that both Anthropic and OpenAI have faced national security scrutiny from the Trump administration and voluntary or forced restrictions on model release. Google, for its part, is "productively engaged with the US government on model testing and broader frameworks." This suggests that even if engineering were faster, regulatory approval may still constrain the timeline. The company's public statement emphasizes speed and cost-effectiveness but does not address the coding gap, signaling that the internal frustration may not fully resolve even after release.

No comments yet. Be the first to share your thoughts!