Google Launches Gemini 3.7 Flash: Its Most Intelligent Workhorse Model Yet for Coding and Agents

From a product positioning perspective, Gemini 3.7 Flash continues the Gemini Flash series' consistent 'high value workhorse model' route, but this time Google places particular emphasis on the two keywords 'agents' and 'coding.' This means the new model's optimization focus is no longer limited to traditional text generation or Q&A, but centers on the model's ability to execute complex tasks as an autonomous agent — including writing and debugging code, calling external tools, planning multi-step workflows, and maintaining task consistency and coherence over long-running sessions. For developers, this is a foundational model built for the 'agent era': it must not only give the right answer in a single conversation, but also be able to complete end-to-end tasks in real working environments, replacing human labor. Google specifically noted that Gemini 3.7 Flash shows significantly better accuracy and output quality than 3.6 Flash when handling tasks that require combining multiple skills, such as scenarios involving code understanding, document retrieval, and data processing simultaneously.

Alongside 3.7 Flash, Google also unveiled Gemini Spark, its new personal AI agent. According to the official introduction, Gemini Spark is a personal AI agent that runs 24/7, designed specifically for Google AI Pro and Ultra subscribers, covering more than 160 countries and regions. This means users can rely on Spark as an always-online personal assistant, letting it continuously handle various tasks in the background: managing schedules, organizing emails, booking services, tracking information, and even proactively discovering and solving potential problems when the user does not initiate a conversation. With 3.7 Flash as its underlying engine, Gemini Spark's reasoning ability, tool-use capability, and long-task execution ability are all guaranteed. Notably, Google emphasized Spark's positioning of 'working for the user' in its launch — marking Google's official entry into the 'personal AI agent' competition pushed by OpenAI, Anthropic and others, attempting to upgrade AI from a passively responding tool into an autonomous system that proactively completes tasks for users.

In terms of tool use, a key upgrade of Gemini 3.7 Flash is the significantly improved support for Google Workspace applications. Google Workspace is one of the most widely used productivity suites in the world, covering Gmail, Google Docs, Google Sheets, Google Slides, Google Calendar and other applications. The improvement in 3.7 Flash's Workspace tool use means the model can operate these applications more accurately and reliably through natural language instructions — for example, automatically composing and sending emails, batch processing data in spreadsheets, generating presentations, and coordinating meeting times. This capability is especially important for Gemini Spark: as a 24/7 personal agent, Spark needs to interact frequently with the user's work tools, and the accuracy and stability of tool use directly determine whether it can truly take on the responsibility of 'working for the user.' Google emphasized that this improvement makes the model more robust when handling complex workflows involving multiple Workspace applications, reducing error-prone calls and context loss.

In terms of pricing, Google has set an attractive introductory price for Gemini 3.7 Flash. The introductory pricing is valid until December 31, 2026; from January 1, 2027, input tokens are priced at $1.50 per million and output tokens at $7.50 per million. This pricing structure continues the Gemini Flash series' consistent 'low input cost, relatively higher output cost' strategy, and also means that during the introductory period users can experience the new model at a more favorable price ahead of time. Looking at Google's overall pricing strategy, the Flash series has always been positioned as the preferred model for 'cost-sensitive scenarios' — suitable for large-scale invocation, batch task processing, and enterprise applications that are sensitive to per-call costs. While maintaining this cost-performance advantage, 3.7 Flash raises performance to a level close to flagship models, making it extremely competitive on the key metric of 'performance/price ratio.' For developers building AI agent applications, a model that is both powerful and cheap means lower operating costs and a more viable business model.