By launching Gemini 3.7 Flash, Google is bringing one of its most exciting upgrades straight to Gemini Spark. Starting today, the cloud-based agent is transitioning to the new model, giving it much better tool-handling skills and a serious performance boost for complex, multi-step tasks.

Tulsee Doshi, senior director and head of Product for Gemini Models, has described the Flash tier as a "workhorse model" that allows companies to easily tackle bulk tasks. With this upgrade, Gemini Spark can now continue working even after a user's laptop is closed or their phone is locked.

Spark Gets a Major Capability Boost

Across Google Workspace, Spark can effortlessly operate across services such as Gmail, Drive, Docs, Calendar, Keep, and Tasks to consolidate files, draft emails, and update status documents. Google has also pushed these abilities further through Google Chrome.

Thanks to this update, the assistant can tap into browser-saved passwords and signed-in accounts to seamlessly browse sites. It can now autonomously compare flight options, schedule apartment viewings, and begin bookings. Before executing anything sensitive like a final payment, however, the system wisely hands the reins back to the user.

To keep these workflows running smoothly, Gemini 3.7 Flash focuses heavily on reliability. Google notes that the model spends extra effort planning and using tools, adapts much better when it hits a roadblock, and ultimately needs fewer retries and less manual intervention.

In related updates, Google is also enhancing how Gemini handles fact-seeking prompts by displaying links to related content, accessible by clicking a gray arrow at the end of a paragraph. Additionally, Gemini is gradually rolling out in Google Messages to users in the European Economic Area, the UK, and Switzerland, while Gemini for teens will be available in over 40 languages.

Putting the Numbers to the Test

Recent benchmark tests firmly back up Google's performance claims. Hitting a 30.4% score on AutomationBench, Gemini 3.7 Flash showed a massive leap over the 17% earned by Gemini 3.6 Flash. It also left rivals in the dust, easily beating GPT-5.6 Terra's 23.6% and Claude Sonnet 5's 10.7%.

When handling document-heavy jobs, the new Flash model scored 34% on GDP.pdf. This comfortably passed Gemini 3.6 Flash at 22%, Claude Sonnet 5 at 28%, and GPT-5.6 Terra at 24.7%.

Coding benchmarks saw similar gains. Edging out Claude Sonnet 5's 42.7% and GPT-5.6 Terra's 41.3%, Gemini 3.7 Flash grabbed a 43.6% on FrontierCode. It simultaneously showed clear improvements over its predecessor on the DeepSWE benchmark.

Undercutting the Market

Hoping to win over developers, Google is rolling out the model with deeply discounted introductory rates that last until December 31. Running Gemini 3.7 Flash will cost just $0.75 per million input tokens and $3.75 per million output tokens.

That pricing makes it a significant bargain compared to Claude Sonnet 5, which charges $2 and $10, and GPT-5.6 Terra, which costs $2 and $12. This aggressive pricing strategy comes as Google already processes over 1 trillion tokens per day on its API amid a fierce performance war with OpenAI.

Even though this Flash update brings a fantastic boost to Spark, power users are still tapping their feet waiting for the more ambitious Gemini 3.5 Pro. Originally previewed at Google I/O in May, that higher-end version still hasn't shipped after missing its expected June rollout.