AI news · Google · Gemini
Google staff doubt Gemini 4 on real coding work, Bloomberg says
Employees say the new model shines on benchmarks but stumbles on front-end code, on the day of its release; Google calls that inaccurate.
Some Google employees doubt that Gemini 4 matches its benchmark scores on real coding work, Bloomberg reported on Wednesday 30 September 2026, the day Google released Gemini 4 Argon. Staff told the outlet the model struggles most with front-end design.
Opinion inside the company is split. Some employees think Anthropic’s and OpenAI’s newest models are improving faster, while others say Gemini 4 has caught up. Google had already dropped a planned Gemini 3.5 Pro release in June. A spokesperson said it would be inaccurate to describe Gemini 4 as underperforming in coding, and DeepMind’s Koray Kavukcuoglu told Bloomberg he is certain Google stays at the frontier.
Investors noticed. Alphabet shares, up more than 2% earlier in the session, were up only about 0.5% heading into the close after the report, Investing.com reports. Gemini models sit behind Search, Maps, Gmail and Chrome, products that each serve more than a billion users, the same report notes.
Google released Argon to cyber defenders first, as we reported on 1 October.
For teams choosing a coding model, the report is a reason to judge Gemini 4 on their own repositories. Google’s own staff say leaderboard scores miss front-end work.
Earlier on this story: Google releases Gemini 4 Argon to cyber defenders first
Sources
- Bloomberg (via Yahoo Finance): Google Grapples With Employee Skepticism About New Gemini Model. Checked
- Investing.com: Alphabet stock slips on report of internal doubts over Gemini 4. Checked
Alexandre S.
Alexandre S. is a creative strategist and copywriter with ten years in content and SEO. He writes and edits short-form ads, works in English and French, and reads the pricing and terms pages behind every tool on this site before anything is written about it.
More AI news from
- FTC opens a safety probe into OpenAI and Anthropic
- California subpoenas OpenAI as agents probed 55 websites
- OpenAI fires three safety researchers over shared information
- Anthropic moves its IPO towards mid November
- Reddit ends RSS feeds and public API access over AI scraping
- ChatGPT adds virtual try-on to its shopping results
- Google puts its first AI chip satellite in orbit
- America.gov chatbot stops answering 2020 election questions
- Time reports Trump asked Grok before the Maduro raid