Google drops Gemini 3.7 Flash model, and it’s ready to handle your chores with the Spark agent

Google has launched Gemini 3.7 Flash, and one of its biggest upgrades is going straight to Gemini Spark.

The company says Spark is moving to the new model starting today, giving its cloud-based agent better tool use and stronger performance on multi-step tasks. Google specifically points to work such as consolidating files, drafting emails, and updating status documents across Google Workspace.

Spark could get a lot better at doing things for you

Gemini Spark can already continue working after a laptop is closed or a phone is locked, and it can operate across services such as Gmail, Drive, Docs, Calendar, Keep, and Tasks.

Google recently pushed those abilities further through Chrome. Spark can use accounts you are already signed into and passwords saved in the browser to navigate websites, compare flight options, schedule apartment viewings, and begin bookings. It still hands control back before sensitive actions such as payments.

Gemini 3.7 Flash is supposed to make those workflows more reliable. Google says the model spends more effort on planning and tool calls, adapts better when it runs into roadblocks, and should need fewer retries and less manual intervention.

The benchmark results point in the same direction. Gemini 3.7 Flash scored 30.4% on AutomationBench, up from 17% for Gemini 3.6 Flash. It also came in ahead of GPT-5.6 Terra at 23.6% and Claude Sonnet 5 at 10.7%.

Gemini 3.7 Flash is stronger beyond Spark too

Google also posted a sizable jump on document-heavy 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.

The pricing is what makes those gains more interesting. Gemini 3.7 Flash costs $0.75 per million input tokens and $3.75 per million output tokens under introductory pricing through December 31. It is significantly cheaper than Claude Sonnet 5 at $2 and $10, and GPT-5.6 Terra at $2 and $12.

The model also posted improvements on coding-focused benchmarks. Gemini 3.7 Flash scored 43.6% on FrontierCode, edging out Claude Sonnet 5 at 42.7% and GPT-5.6 Terra at 41.3%, while also improving over Gemini 3.6 Flash on DeepSWE.

For Spark, Gemini 3.7 Flash looks like a meaningful upgrade, especially if the improvements in tool use and multi-step planning translate into fewer failed tasks and less manual intervention. Power users, however, are still waiting for Google’s more ambitious Gemini 3.5 Pro model, which was previewed at Google I/O in May but has yet to ship after missing its expected June rollout.