The Zhitong Finance App learned that Alphabet (GOOGL.US, GOOG.US) shares fell 4% on Thursday due to news of the delay in the release of the flagship artificial intelligence model. According to media reports, since the model's performance has not met internal expectations, Google decided to postpone the launch of the Gemini 3.5 Pro large language model, which is several months behind the original plan.
According to people familiar with the matter, the development progress of Gemini 3.5 Pro lags behind expectations, mainly because Google continues to optimize model performance. Among them, code generation capabilities have fallen short of the company's internal goals, and this field is becoming one of the most competitive racetracks for global AI vendors.
The report said that at a time when Google is slowing down the pace of product launches, OpenAI and Meta Platforms (META.US) have recently launched next-generation AI models. Their code generation capabilities are generally superior to Google's existing products, further intensifying market competition.
Google first unveiled the Gemini 3.5 Pro at the Google I/O developer conference in May of this year, and stated that the model has already begun to be used within the company and was originally planned to be made available to more users in the following month. However, the latest news shows that the plan has been delayed.
In response to related reports, an Alphabet spokesperson said that the company continues to rapidly launch various AI models while maintaining high cost efficiency. The spokesperson said that the company is currently testing Gemini 3.5 Pro, upgraded Flash models, and other AI models with partners, and is actively cooperating with the US government.
In recent years, code generation has become one of the most important application scenarios for generative AI, and it is also the focus of competition among major AI model providers. Vendors including OpenAI and Anthropic continue to launch code generation models for developers, and some of these open source models are free for developers to use.
Just last week, Meta released Muse Spark 1.1, a next-generation big language model. Alexander Wang, chief AI officer of Meta, said that this is the company's most powerful model in the field of AI agents and code generation so far.
Also last week, OpenAI launched the GPT-5.6 Sol model. OpenAI CEO Sam Altman said that the new model increased the efficiency of using tokens in AI agent code generation tasks by 54%, further strengthening the model's competitive advantage between performance and cost.
Analysts believe that as AI model competition gradually shifts from parameter scale to actual application capabilities and cost efficiency, code generation ability has become an important indicator for measuring the competitiveness of large language models. The delayed release of Gemini 3.5 Pro also reflects Google's desire to officially launch the model after its performance is mature to cope with increasingly fierce competition in the industry.