Gemini-SQL2: natural language becomes executable SQL queries

What has changed: Google Research redefines text-to-SQL

On June 13, 2026, Google Research announced Gemini-SQL2 , an advanced text-to-SQL conversion system. The model is built on Gemini 3.1 Pro and represents a qualitative leap compared to previous generations. According to reports from The Decoder , Gemini-SQL2 has reached the 80.04% accuracy on the BIRD benchmark, the industry standard for evaluating text-to-SQL systems.

This result clearly beats out competing models from OpenAI and Anthropic. So, Google is stepping up as the tech leader in this specific area of AI applied to data. The BIRD benchmark measures how well a system can create accurate SQL queries from everyday questions on real, complex databases.

Plus, Google said that the tech behind Gemini-SQL2 will be built into its existing data services. These include BigQuery, Looker, and other Google Cloud tools. So, companies already using the Google ecosystem can enjoy these features without needing new tools.

The BIRD benchmark and the meaning of 80% accuracy

The BIRD benchmark (Big Bench for Large-scale Database Grounded Text-to-SQL Evaluation) is the go-to industry standard. It tests models on real-world databases with tricky schemas and vague questions. Hitting 80% on this test is no walk in the park.

In fact, older systems scored significantly lower. The gap between Gemini-SQL2 and its competitors is, according to published data, several percentage points wider. Still, we need to keep things in perspective: that remaining 20% error rate can still lead to broken or incomplete queries in live environments. So, having a human double-check things is still a must for high-stakes scenarios.

To dive deeper into the technical workings of AI benchmarks, you can refer to the analyses published by MIT Technology Review , which has covered the topic of evaluating language models on structured tasks several times.

Immediate impact for Italian SMEs without a technical team

For many Italian B2B and retail SMEs , access to company data is still mediated by technical roles. A sales manager who wants to know which customers bought twice in the last quarter has to wait for a developer to write the query. This slows down decision-making.

Gemini-SQL2 could shake things up. So, a system that turns questions in Italian — or English — straight into runnable SQL really lowers the tech barrier. Plus, the 80% accuracy on the BIRD benchmark shows the system works great even with tricky database structures.

Similarly, tools like these integrate with the trends of Business applied AI which we at SHM Studio follow for our clients. In particular, automated data access is one of the most concrete, high-return use cases for mid-sized companies. For this reason, it is worth evaluating how this technology fits into existing operational workflows.

How it integrates with the Google Cloud ecosystem

Google has announced that Gemini-SQL2 will be rolled out within its cloud services. BigQuery , Google Cloud's data warehouse, is the most obvious candidate for an initial integration. Looker, the business intelligence platform acquired by Google, could benefit even more directly.

As a result, small and medium businesses that have already jumped into the Google ecosystem could get these features through step-by-step updates to the tools they already use. Still, the official launch dates haven't been shared yet. So, it's way too early to map out hands-on integrations anytime soon until we hear more.

For those managing campaigns and data on Google platforms, it is also worth keeping an eye on developments related to Google Ads and its automated reporting features, which could indirectly benefit from this technology. Similarly, anyone working with Digital marketing data-structured will be interested in following the evolution of these tools.

The competitive landscape: OpenAI and Anthropic are lagging behind on this specific task

Gemini-SQL2's achievement is a big deal because it drops right when OpenAI and Anthropic are hogging the AI spotlight. Still, the BIRD benchmark tells a different story for this specific area.

OpenAI's models — including GPT-4o and recent versions — and Anthropic's such as Claude 3.7 haven't reached comparable levels on structured text-to-SQL. Thus, Google shows that vertical specialization on a specific task can yield measurable competitive advantages. This is a significant strategic signal for the market.

According to the analyses of Gartner regarding tech trends, specializing AI models for specific industries is one of the most exciting paths for 2026-2027. So, Gemini-SQL2 fits right into a broader trend of technical competition among the big players.

What to do now: three operational considerations for SMEs

First of all, it is helpful to map existing company databases and identify which queries are most frequently requested of technical staff. This exercise helps estimate the potential time savings a text-to-SQL tool could generate.

Next, it's worth checking out if your company's data setup is already on Google Cloud or if you're planning a move. Actually, connecting Gemini-SQL2 will be a breeze for anyone already in the Google world. But if you're still using on-premise databases or other cloud providers, getting this tech up and running might take a few extra steps.

Lastly, don't wait for the official release to start getting your data in order. Good data structure is a must-have for any natural language querying tool. We at SHM Studio we can support SMEs in this assessment and planning phase, as part of our services for AI consulting and digital marketing data-driven .

What press releases don't say: real limits to consider

Benchmarks are helpful, but they don't tell the whole story. The 80.04% accuracy on BIRD is a result in a controlled environment. In real business settings, databases have non-standard column names, implicit relationships, and undocumented internal conventions. Therefore, actual performance might be lower than what is measured in the lab.

Also, the quality of the generated queries really depends on how well you ask the question. A non-technical user might ask vague things, leading to messy results even with a top-notch model. So, training end users is still super important for these tools to succeed.

Despite this, the direction is clear. The gap between natural language and structured data is shrinking at an accelerated pace. For Italian SMEs, this means that investing today in the quality and organization of their data — through services like web development data-driven or strategies of SEO based on structured analytics — is a strategic choice with an increasingly shorter payback horizon.

Outlook: where this technology will lead in 2027-2028

In the short term, the integration of Gemini-SQL2 into Google Cloud products represents the most likely evolution. In the medium term, similar features are expected to become standard across major business intelligence platforms, from Tableau to Power BI.

Therefore, by 2027-2028, querying company databases in natural language could become a basic feature rather than a differentiator. As a result, competitive advantage will shift from the ability to access data to the ability to interpret and act on it quickly.

To dive deeper into AI applied to Italian SMEs, a helpful read is available on SHM Studio blog , where we publish regular analyses on these topics. Anyone wanting to discuss how to integrate AI tools into their business processes can contact us directly . Also, for those managing content and digital communication, our services for SEO copywriting and LinkedIn campaigns they can support the storytelling of these changes towards customers and stakeholders.