Google has released the Gemini 3.7 Flash model, which is equipped to assist you with your tasks using the Spark agent.

      Google has introduced Gemini 3.7 Flash, and one of its key enhancements is aimed at improving Gemini Spark.

      The company states that Spark will transition to this new model starting today, providing its cloud-based agent with improved tool usage and better performance on multi-step tasks. Google highlights tasks such as file consolidation, email drafting, and updating status documents within Google Workspace as specific areas of enhancement.

      Spark has the potential to significantly improve its ability to assist users.

      Gemini Spark is already capable of continuing its tasks even when a laptop is closed or a phone is locked, and it can function across various services including Gmail, Drive, Docs, Calendar, Keep, and Tasks.

      Recently, Google expanded these capabilities via Chrome. Spark can utilize already signed-in accounts and saved passwords to browse websites, compare flight options, arrange apartment viewings, and initiate bookings, while still returning control before sensitive actions like payments.

      Gemini 3.7 Flash aims to enhance the reliability of these workflows. Google indicates that the model invests more effort into planning and tool utilization, adapts more effectively when encountering obstacles, and requires fewer retries and less manual intervention.

      Benchmark results also support this improvement. Gemini 3.7 Flash achieved a score of 30.4% on AutomationBench, up from 17% for Gemini 3.6 Flash. Additionally, it outperformed GPT-5.6 Terra, which scored 23.6%, and Claude Sonnet 5, which scored 10.7%.

      Moreover, Gemini 3.7 Flash excels in document-intensive tasks as well. It scored 34% on GDP.pdf, in contrast to 22% for Gemini 3.6 Flash, 28% for Claude Sonnet 5, and 24.7% for GPT-5.6 Terra.

      The pricing structure adds further interest to these advancements. Under introductory pricing lasting until December 31, Gemini 3.7 Flash is priced at $0.75 per million input tokens and $3.75 per million output tokens, making it considerably cheaper than Claude Sonnet 5, which costs $2 and $10, and GPT-5.6 Terra, priced at $2 and $12.

      Improvements have also been noted in coding-related benchmarks. Gemini 3.7 Flash scored 43.6% on FrontierCode, surpassing Claude Sonnet 5 at 42.7% and GPT-5.6 Terra at 41.3%, while also performing better than Gemini 3.6 Flash on DeepSWE.

      For Spark, Gemini 3.7 Flash appears to be a significant upgrade, particularly if the enhancements in tool usage and multi-step planning lead to fewer task failures and reduced manual intervention. Nevertheless, power users are still anticipating Google’s more advanced Gemini 3.5 Pro model, which was previewed at Google I/O in May but has yet to be released after missing its expected June launch.