Zapier Adds Gemini 3.7 Flash as AutomationBench’s New Leader.

Zapier has added Google’s Gemini 3.7 Flash to its automation platform, pitching the model as a rare combination of top benchmark performance and comparatively low operating costs. The headline figure is a 30.4% score on AutomationBench, which Zapier says puts the model in first place.

The company announced the availability on LinkedIn, alongside claims that Gemini 3.7 Flash costs less than half as much as GPT 5.6 Sol and Opus 5. It also reportedly finished 11.7 percentage points ahead of every other tested model on finance workflows.

For small businesses building AI into operational automations, that cost-to-performance balance may matter more than raw model intelligence. A model that is slightly cheaper per run can produce meaningful savings when it is processing invoices, qualifying leads, categorising support requests or transforming records thousands of times each month.

AutomationBench is Zapier’s benchmark for evaluating how effectively AI models complete workflow-oriented tasks. According to the company’s benchmark scorecard, Gemini 3.7 Flash reached 30.4%, becoming the new overall leader.

That result should still be interpreted carefully. A 30.4% score does not mean the model will complete 30.4% of every company’s automations correctly, nor does it establish that Gemini is the best model for every prompt or application. Benchmarks measure performance against a defined collection of tasks, scoring rules and operating conditions. Real-world results will also depend on prompt design, available context, tool permissions and the quality of incoming data.

The finance result is particularly notable. Zapier says Gemini 3.7 Flash led competing models by 11.7 points in that workflow category, suggesting a potentially meaningful advantage for automations involving financial documents or structured business information. The announcement does not provide task-level results, however, so builders should test the model against their own invoices, reports and approval rules before putting it into a consequential workflow.

The leaderboard is also moving quickly. Zapier previously added Opus 5 after its own AutomationBench sweep, but Gemini 3.7 Flash has now displaced it at the top of the overall ranking. That volatility is another reason for businesses to avoid designing workflows around the assumption that one model will remain the strongest indefinitely.

The immediate benefit is another model choice inside Zapier’s automation ecosystem. The company’s guide to available AI models lists the major providers that can be connected to Zap workflows and outlines the jobs for which each is best suited.

Builders considering Gemini 3.7 Flash should focus on controlled comparisons rather than automatically replacing a working model. A practical evaluation could include:

• Running the same representative inputs through the current model and Gemini 3.7 Flash
• Comparing accuracy, latency and total per-run cost
• Checking structured outputs for missing or incorrectly formatted fields
• Requiring human approval for financial or customer-facing actions during testing
• Keeping a fallback model available for errors or degraded responses

The release also supports Zapier’s broader push toward AI model flexibility. For smaller teams, the strategic value is not simply access to another model. It is the ability to switch as benchmark performance, pricing and business requirements change.

Zapier did not provide exact token pricing, eligible subscription plans or detailed rollout conditions in its LinkedIn announcement. Customers should therefore confirm costs and availability within their own accounts before estimating savings.

What is Gemini 3.7 Flash in Zapier?

Gemini 3.7 Flash is a Google AI model now available for use in Zapier’s automated workflows. It can serve as the model behind AI-powered steps, with Zapier positioning it as a lower-cost option that also performs strongly on its AutomationBench evaluation.

How did Gemini 3.7 Flash perform on AutomationBench?

Zapier reports that the model scored 30.4%, ranking first overall at the time of the announcement. It also led every other tested model on finance workflows by 11.7 percentage points.

When is Gemini 3.7 Flash available, and what does it cost?

Zapier announced the model as available on August 13, 2026. The company says it costs less than half as much as GPT 5.6 Sol and Opus 5, but the announcement does not specify exact prices or required Zapier plans.