Google has launched Gemini 3.7 Flashand one of its biggest updates is going straight to the Gemini Spark.
According to the company, Spark will switch to the new model starting today, which will provide better resource utilization and stronger performance for its cloud-based agent for multi-step tasks. Google specifically points to tasks like merging files, drafting emails, and updating status documents in Google Workspace.
Spark could do a lot better for you
Gemini Spark can continue working even after you close your laptop or phone, and can work between services like Gmail, Drive, Docs, Calendar, Keep, and Tasks.
Google recently enhanced these capabilities through Chrome. Spark can use the accounts you’re already signed in to and the passwords you’ve saved in your browser to navigate websites, compare flight options, schedule home viewings, and start bookings. It still returns control before sensitive operations like payouts.
Gemini 3.7 Flash is said to make these workflows more reliable. According to Google, the model puts more effort into design and tool calls, adapts better when it hits roadblocks, and requires less retries and manual intervention.
The benchmark results point in the same direction. Gemini 3.7 Flash scored 30.4% on AutomationBench, compared to 17% for Gemini 3.6 Flash. It also beat the GPT-5.6 Terra by 23.6% and the Claude Sonnet 5 by 10.7%.
Gemini 3.7 Flash is even more powerful than Spark
Google also posted a major jump in document-rich work. Gemini 3.7 Flash scored 34% on GDP.pdf, compared to 22% for Gemini 3.6 Flash, 28% for Claude Sonnet 5, and 24.7% for GPT-5.6 Terra.
Pricing is what makes these gains more interesting. Gemini 3.7 Flash will cost $0.75 per million input tokens and $3.75 per million output tokens, according to introductory prices valid until December 31st. It’s significantly cheaper than the Claude Sonnet 5 at $2 and $10 and the GPT-5.6 Terra at $2 and $12.
The model also released improvements to benchmarks focused on coding. Gemini 3.7 Flash scored 43.6% on FrontierCode, ahead of Claude Sonnet 5 with 42.7% and GPT-5.6 Terra with 41.3%, while also outperforming Gemini 3.6 Flash on DeepSWE.
For Spark, Gemini 3.7 Flash seems like a sensible upgrade, especially if improvements in tooling and multi-step design result in fewer failed tasks and manual intervention. But power users are still waiting for Google’s more ambitious Gemini 3.5 Pro, which was unveiled at Google I/On in May but has yet to ship, missing the expected June release.
\
