Google's newest artificial intelligence model is out, but it has not silenced the doubters. The company announced Gemini 4, its flagship model, on Sept. 30 after months of delays, Reuters reported. Google first made the model, code-named Argon, available to select cybersecurity partners rather than to everyone at once, Reuters reported.

Inside Google, some workers are not convinced. The Northwest Arkansas Democrat-Gazette, in a report on Sunday, described skepticism among employees about Gemini 4's ability to write and fix computer code, one of the most valuable uses of AI for businesses.

Coding has become a key battleground. Companies pay for AI tools that can help programmers build software faster. Rivals such as OpenAI and Anthropic have pushed hard into that market, and their tools are widely used by developers.

Why release first to security partners? Powerful models can help defenders find software flaws, but they can also help attackers. A slower rollout lets Google test the model with experts before wider use. It also gives the company time to fix problems.

Here is the background. Google invented the "transformer" technology behind most modern AI models in 2017, but it was caught off guard when OpenAI released ChatGPT in late 2022. Since then, Google has raced to catch up, merging its AI labs into Google DeepMind and folding Gemini into Search, Gmail and Android phones.

For everyday users, Gemini 4 will likely show up over time in Google's apps and services. The company has not said when the model will be widely available.

The pressure is high. Google's parent, Alphabet, is spending tens of billions of dollars a year on data centers and chips to power AI.

What we know: Gemini 4 launched Sept. 30 to select partners after delays.
What we don't: How well it performs in wide use, and when ordinary users will get it.