Google's Gemini 3.5 Pro launch delayed again due to coding performance issues

Google's highly anticipated AI model, Gemini 3.5 Pro, is reportedly experiencing significant launch delays, extending months beyond its initial schedule due to persistent performance problems, especially in coding tasks. This setback fuels concerns that Google may be losing ground in the intense artificial intelligence race against competitors like OpenAI and Anthropic.

Sources within Google, including current and former employees, have indicated growing frustration among engineers, researchers, and managers over the repeated postponements. Alphabet CEO Sundar Pichai had announced a June release for Gemini 3.5 Pro at the company's annual developer conference in May. However, as of mid-July, a concrete launch date remains unconfirmed, with the company citing ongoing performance improvements and testing.

The primary obstacle appears to be the model's coding performance. Despite Google updating Gemini's training data last month to enhance its coding abilities, the results have reportedly fallen short of internal benchmarks. This news has impacted Alphabet's stock, which fell 4.4% in pre-market trading following the reports.

“

We are working to launch a variety of AI models quickly while maintaining cost efficiency.

Beyond technical hurdles, the complexity of Google's organizational structure and decision-making processes are also cited as contributing factors to the delays. Reports suggest that while top executives, including co-founder Sergey Brin, are pushing for faster development and release to secure a leading position in the AI coding market, different divisions like Google Cloud, Google DeepMind, and Android are independently developing their own AI coding tools. This lack of synergy and conflicting interests are hindering progress. One former employee likened the effort to align leadership across departments to 'boiling the ocean.' Some engineers are also reportedly resistant to rapid model development, preferring a more cautious approach to coding.

In response to the reports, Google stated, 'We are working to launch a variety of AI models quickly while maintaining cost efficiency.' The company confirmed it is testing Gemini 3.5 Pro and the enhanced Flash model, while also engaging with the U.S. government on model verification and AI safety frameworks.

“

It is as difficult as boiling the ocean to get leadership across all departments to move in one direction.