Google delays Gemini 3.5 Pro, internal conflicts and competition heat up

Google's flagship AI model Gemini 3.5 Pro has been delayed for months, as the company wants to improve its performance, especially in coding. This has frustrated engineers, researchers, and managers internally.

Competitors like Anthropic and OpenAI keep releasing better models, especially for coding, making Google's existing products look outdated. Internal politics and the need to integrate AI into products like Search and Maps have also slowed things down.

Google tried swapping some training data to boost coding skills, but tests didn't meet expectations. There's also a struggle inside Google over who leads AI coding tools, with some engineers worried about using AI for important code. The company is trying to unify its tools, but engineers often face capacity issues.

Some researchers have left for Anthropic or other labs. Google now limits access to Anthropic's Claude model to only a few teams working on top-priority projects.