Alphabet drops after report says Gemini 3.5 Pro rollout is delayed

Alphabet shares fell Thursday after Bloomberg reported that Google has delayed the broader release of Gemini 3.5 Pro, its most powerful artificial intelligence model. The move matters for investors because Google’s AI progress is a key part of the case for Alphabet’s growth as rivals race to win developers and business customers.

Alphabet’s Class A shares were down about 4% Thursday, according to CNBC. Bloomberg reported, citing people familiar with the matter, that Gemini 3.5 Pro is months behind schedule while Google works to improve the model’s performance.

An AI model is software trained on large amounts of data to produce text, code, images or other outputs in response to prompts. For Google, Gemini is the brand at the center of its push to compete with OpenAI, Meta, Anthropic and other companies building advanced AI systems.

What the reported delay is about

Bloomberg reported that Gemini 3.5 Pro’s coding abilities have fallen short of Google’s internal targets. Coding has become one of the most closely watched AI uses because developers can use these systems to write, review and fix software faster.

The timing is sensitive. Bloomberg said OpenAI and Meta have recently released AI models that beat Google’s current offerings at generating software code. That puts extra attention on whether Google can turn its research strength into products that developers and companies choose to use.

Google announced Gemini 3.5 Pro in May at its annual Google I/O developer conference. At that time, the company said the model was already being used inside Google, but would not be ready for a wider release until the next month, according to CNBC.

Google says it is still testing models

An Alphabet spokesperson told CNBC by email that the company is “shipping quickly across a wide range of models while keeping them highly cost-effective for customers.”

“We’re currently testing 3.5 Pro, an upgraded Flash model, and other models with partners, and we’re productively engaged with the U.S. government,” the spokesperson said.

Flash is Google’s name for a faster, lighter Gemini model line. In AI, cost can matter almost as much as raw performance because customers often pay based on usage. More efficient models can make frequent AI tasks cheaper to run.

Why coding models are getting so much attention

AI coding tools have become a major battleground for model providers. CNBC reported that companies including Anthropic and OpenAI, along with Chinese AI labs such as Z.ai, are offering models aimed at software development. Some Chinese labs have released open-weight variants, meaning developers can access model components through the open-source ecosystem.

Meta introduced Muse Spark 1.1 last week. Meta AI chief Alexandr Wang called it the company’s “strongest model for agentic and coding work yet,” according to CNBC. Agentic coding refers to AI systems that can handle multi-step software tasks with less direct instruction from the user.

OpenAI also released GPT-5.6 Sol last week. CEO Sam Altman said the model is 54% more token efficient on agentic coding tasks, CNBC reported. Tokens are the chunks of text a model processes, so token efficiency can affect how much a task costs and how quickly it runs.

For Alphabet shareholders, the report adds a clear test: Google has to show that its next Gemini release can compete on performance, cost and timing in one of AI’s most commercially important categories.

This story draws on original reporting from CNBC.