Google’s flagship AI model, Gemini 3.5 Pro, has yet to receive a public release despite being announced at Google I/O with an expected launch in June. Multiple reports suggest the company has postponed the model several times as engineers work to improve its coding capabilities and performance on complex AI tasks.

While Google has not publicly confirmed the reasons for the delay, reports indicate that the company has been refining the model’s architecture and testing it with select enterprise customers before a wider rollout.

The delay comes at a time when competition in AI coding assistants is intensifying, with rivals continuing to release increasingly capable models for software development.

What Happened to Gemini 3.5 Pro?

At Google I/O 2026, Google introduced the Gemini 3.5 family of AI models.

The company released Gemini 3.5 Flash, its faster, more lightweight model, and announced that the more capable Gemini 3.5 Pro would follow in June.

That public release has yet to occur.

Although Gemini 3.5 Pro has remained unavailable to most users, Google has confirmed that it is being tested with selected partners before a broader launch.

Why Has the Release Been Delayed?

Google has not publicly provided a detailed explanation for the delayed launch.

However, reports from outlets including Bloomberg have suggested that engineers have continued refining the model’s performance, particularly in areas such as:

• Software coding.
• Multi-step reasoning.
• Token efficiency.
• Tool use and autonomous workflows.

According to those reports, Google has extended internal development after testing revealed shortcomings in these areas.

The company has not independently confirmed every reported detail regarding its internal development process.

Reports of a Model Rebuild

Some reports have suggested that Google opted to make significant architectural changes to Gemini 3.5 Pro after earlier improvements failed to deliver the desired performance.

According to these reports, engineers encountered challenges involving complex tool-calling workflows—often referred to as “agentic” capabilities—where AI systems autonomously perform sequences of tasks such as writing, testing, and debugging code.

Neither Google nor its researchers have publicly confirmed reports describing a complete ground-up rebuild of the model.

What Is Agentic AI?

Agentic AI refers to systems capable of carrying out multi-step tasks with limited human intervention.

Instead of simply answering a question, an agentic model may:

• Write software.
• Execute code.
• Search documentation.
• Analyze results.
• Revise its work.
• Complete an entire workflow automatically.

Reliable agentic behavior has become one of the most competitive areas of AI development because it allows models to function as software assistants rather than simple chatbots.

Why Is Coding So Important?

Coding has become one of the most valuable commercial applications of generative AI.

Developers increasingly rely on AI systems to:

• Generate code.
• Identify bugs.
• Explain unfamiliar programming languages.
• Build applications.
• Automate repetitive engineering tasks.

Because of this, benchmark performance on programming tasks has become a closely watched measure of an AI model’s capabilities.

Companies compete aggressively to improve coding performance, as enterprise customers often evaluate models based on real-world software development tasks rather than conversational ability alone.

How Is Google Responding?

Google has stated that it is continuing to test Gemini 3.5 Pro with selected partners.

According to the company’s public statements, testing currently includes:

• Enterprise customers.
• Strategic partners.
• Government organizations.
• External evaluators.

Google has also indicated that it is evaluating updated versions of Gemini Flash alongside Gemini 3.5 Pro.

The company has not announced a revised public release date.

What Does This Mean for Google’s AI Strategy?

Google remains one of the world’s leading AI companies, having pioneered the transformer architecture that underpins many modern large language models.

However, competition has accelerated rapidly.

Companies including OpenAI, Anthropic, xAI, Meta, and several Chinese AI developers continue releasing increasingly capable models, particularly in software engineering and autonomous coding.

Any delay to Gemini 3.5 Pro gives competitors additional time to strengthen their position in one of AI’s fastest-growing markets.

Why This Matters

The postponement of Gemini 3.5 Pro illustrates how difficult it has become to develop state-of-the-art AI systems.

Modern foundation models are evaluated not only on language understanding but also on reasoning, planning, coding accuracy, tool use, and reliability in enterprise environments.

Rather than releasing a model prematurely, companies increasingly appear willing to delay launches until they believe performance meets commercial expectations.

Whether Google’s additional development time results in a significantly stronger model will become clear once Gemini 3.5 Pro is released more broadly.

The Bottom Line

Gemini 3.5 Pro remains one of the most anticipated AI releases of 2026. While reports suggest Google has postponed the model several times to improve coding and agentic AI performance, the company has officially stated only that the model is still undergoing testing with selected partners.

Until Google announces a public release date, questions about the extent of the reported delays and internal redesign remain based on media reporting rather than confirmed company disclosures.

TL;DR

• Google announced Gemini 3.5 Pro at Google I/O with an expected June release.
• The model has not yet been publicly launched.
• Reports suggest Google delayed the release multiple times while improving coding performance and agentic AI capabilities.
• Gemini 3.5 Pro is currently being tested with selected partners, according to Google.
• The delay highlights growing competition in AI models designed for software development.