Google’s Gemini Pro slips as coding upgrade misses goals

Google is reportedly months behind schedule on the next version of Gemini Pro, with the delay tied to coding performance that has failed to meet internal goals. According to Bloomberg, which cited 10 current and former employees, Google had been widely expected to unveil the upgrade at its May developer conference but has not closed the gap with Anthropic and OpenAI in code generation.

The report says Google updated Gemini’s training data late last month to improve coding ability, but one source told Bloomberg the results were disappointing. Pressure has only grown since OpenAI and Meta released newer models that reportedly outperform Google’s current systems for writing code. After the news, Alphabet shares fell more than 3%.

A Google spokesperson told Bloomberg that the company is “shipping quickly across a wide range of models” and is testing the upgraded Pro, a new Flash model, and other models with partners.

Part of the holdup appears to be internal. Google Cloud, DeepMind, the Android team, and consumer product groups are all building AI coding tools, creating overlapping efforts and competition inside the company. Sergey Brin has reportedly pushed for faster progress in AI coding, but former employees said those efforts have been slowed by rival factions and by engineers who believe important code should still be written by humans to meet Google’s standards.

Google has started consolidating those efforts. Chief AI Architect Koray Kavukcuoglu is working to unify internal coding tools, while a new DeepMind team led by research engineer Sebastian Borgeaud has been set up to focus on the issue.

Recommended reading

Linus Torvalds backs AI coding in Linux

At its latest Cloud conference, Google said 75% of code at the company is now AI-generated and that most developer tooling has been consolidated under Antigravity, its internal platform for managing data, memory, and safety protocols for AI applications.

Customer reaction and talent pressure

The delays have also fed senior departures to Anthropic and other labs, with former employees saying frustration over Google’s competitive position is one reason people are leaving. The report adds that engineers inside Google run into compute constraints when trying to use AI tools themselves, reflecting broader competition for capacity that also affects customers.

Access to Anthropic’s Claude is limited internally, with only some Google teams — mainly those doing cutting-edge research — allowed to use it.

Customers, meanwhile, are split on Google’s current Flash model. Rodrigo Davies, a product manager at Figma, said it hit “a sweet spot of speed and quality” for the company’s AI assistant. But Freddy Vega, CEO of Latin American education platform Platzi, said Flash is more expensive and slower than its predecessor while still trailing rivals, and said his team has moved to Anthropic instead.