Google delays the launch of its AI language model Gemini 3.5 Pro again: The promised date for June 2026 has passed, and the company does not provide a new date. According to a Bloomberg report from July 16, 2026, the model’s programming capabilities particularly fall short of internal goals.
New training data misses set goals
Google had already announced the Pro version of Gemini 3.5 at the developer conference I/O in May 2026 and had indicated a market launch for the following month. At the same time, the company released the smaller variant Gemini 3.5 Flash, but the more powerful Pro version was not released. According to Bloomberg, Google updated the training data at the end of June to improve the model’s programming capabilities. Such multi-step coding tasks occur in professional development environments and are becoming increasingly important for high-paying enterprise customers.
However, the results of this revision have been disappointing, according to sources familiar with the process – an assessment that has not been independently verified. Google is more reserved in its public statements: In a statement to 9to5Google, the company said it is currently testing “3.5 Pro, an improved Flash model, and other models” together with partners. It further stated that it continues to deliver models quickly and cost-effectively. Google did not provide a new release date. Thus, it remains unclear when developers will gain access to the more powerful Pro model.
Google faces pressure in the competition for programming AI
The lag in so-called agentic programming AI – systems that autonomously handle multi-step coding tasks – has been a concern for Google for some time. According to reports, CEO Sundar Pichai has previously admitted that his company is “a bit behind” in this area. He cited a lack of developer-side tools that provide valuable training data for such tasks. In April 2026, Bloomberg reported that Google was losing ground in AI-assisted programming compared to its competitors.
Current and former employees report growing concern that Google could continue to lose ground compared to Anthropic and OpenAI. Both competitors have updated their coding models multiple times in recent months and are now considered benchmarks for demanding programming tasks by many developers. The delayed release of Gemini 3.5 Pro is unlikely to reduce this gap, especially since the competition has continued to improve their models in the meantime while Google is making internal adjustments. For the company, this not only jeopardizes its technological leadership but also market shares among high-paying cloud and enterprise customers who can increasingly choose between multiple providers.
Flash model and predecessor bridge the waiting time
While the Pro version is delayed, Google is simultaneously testing a revised version of Gemini 3.5 Flash with selected partners. The company has not yet provided details on the functionality or performance of the revised Flash model. Google released the original Flash version in May 2026 and positions it as a faster but less powerful and cheaper counterpart to the still-pending Pro version. For many use cases with lower demands for complex reasoning, the Flash variant can already serve as a transitional solution.
Until the release of Gemini 3.5 Pro, Gemini 3.1 Pro from February 2026 remains the company’s latest stable flagship model for demanding tasks. For users and enterprise customers, this means: Those waiting for the announced improvements in complex programming tasks will have to be patient. Meanwhile, Google is working internally on a more convincing version while trying to keep costs low for customers. Developers with production-ready requirements will rely on existing Google models or offerings from competitors until then.
It will be crucial whether Google closes the coding gap before Anthropic and OpenAI further expand their lead in developer tools. The company has not yet provided a binding new target date for Gemini 3.5 Pro.
