Gemini 3.5 Pro Delayed Again, Rumored Internal Performance Inferior to Muse Spark

According to Dynamic Beating monitoring, Google's flagship model Gemini 3.5 Pro has been delayed for several months. Google had originally planned for a June release, but the model's performance has yet to meet internal targets. Even after updating the training data in late June, the results were still unsatisfactory.

APPSO, citing insider sources, suggested that the latest evaluation of the 3.5 Pro might not even outperform Meta's Muse Spark. Some employees with ties to related businesses have recently had their testing permissions revoked, leaving only the model team able to continue using and viewing the evaluation results.

APPSO also reported that morale within the Gemini-related teams is low. Some employees are concerned about team integration and layoffs, leading them to consider joining other AI labs. Recently, Noam Shazeer moved to OpenAI, and John Jumper, a key researcher on AlphaFold, left Google to join Anthropic.

However, there is currently no solid evidence to support the claim that the 3.5 Pro has been canceled. Google responded by stating that the model is still undergoing testing with partners and is also testing upgraded models such as Flash.