Google is running behind on the planned release of Gemini 3.5 Pro, its most capable flagship artificial intelligence model, after internal evaluations found the system fell short of expectations, according to a recent report.
The model had been slated for a June release, a timeline the company’s chief executive announced at its annual I/O developer conference in May. The delay has stretched into months as Google works to address shortcomings highlighted during testing.
Within the company, the setback has prompted concern among engineers, researchers and managers who worry about competing models from other developers that have demonstrated stronger performance, the report said, citing the views of 10 current and former employees.
Google undertook a data refresh late last month intended to improve Gemini’s capabilities, with particular emphasis on coding proficiency. Those updated training data and subsequent results did not deliver the improvements engineering teams had expected, the report added.
- The model was announced for a June release by the CEO at the I/O conference in May.
- Internal unease stems from rival models released by competitors that are considered to outperform Gemini.
- Google updated the model’s training data late last month, but the outcomes remained below internal targets.
Market reaction was swift: shares of Alphabet fell nearly 3% after the information was reported.
"We’re currently testing 3.5 Pro, an upgraded Flash model, and other models with partners, and we’re productively engaged with the U.S. government," a company spokesperson said in a statement.
"We’re shipping quickly across a wide range of models while keeping them highly cost-effective for customers," the spokesperson added.
The broader AI landscape remains active. OpenAI introduced GPT-5.6 last week following a delay that was prompted by requests from the U.S. government about potential national security risks associated with powerful AI technology.
Separately, Anthropic temporarily disabled its most advanced offerings, Mythos 5 and Fable 5, after a June 12 U.S. export control order cited national security concerns. Those restrictions were removed in late June after Anthropic implemented additional safeguards.
These developments underscore an industrywide push to boost model performance, reduce operational costs and expand capabilities for enterprise customers while also responding to heightened regulatory scrutiny. Google says it continues product testing and partner collaborations as it works toward meeting its internal performance goals for Gemini 3.5 Pro.