Google claims Gemini 3.7 Flash delivers improved performance in coding, software engineering, web development and complex multi-step tasks, outperforming its predecessor on several benchmarks.(Image: Google)The race to build more capable AI models is increasingly shifting towards coding and AI agents. On Thursday, August 13, Google released a new version of its Gemini AI model, Gemini 3.7 Flash, that has been designed to handle complex software engineering multi-step planning and enterprise workflows while keeping the cost of running AI agents low.Gemini 3.7 Flash, the latest addition to its Flash family of AI models, is being described by the company as its ‘most intelligent workhorse model yet’ for coding and AI agents. The model is being launched just three weeks after the arrival of Gemini 3.6 Flash. While AI models are designed primarily to generate text or answer queries, Gemini 3.7 Flash leans heavily towards coding, software engineering, web development, and agentic workflows. According to Google, the model is better at understanding instructions, planning multiple steps, using tools and adapting when it encounters problems. The model reportedly is aimed at reducing the amount of manual intervention that developers need when working with AI agents.How is Gemini 3.7 Flash different?Perhaps the most noticeable change is the model’s focus on complex workflows. Google said that Gemini 3.7 Flash improves on its predecessor when it comes to tasks like debugging, resolving software issues, and generating production-ready code. On the software-engineering benchmarks cited by Google, Gemini 3.7 Flash scored 43.6 per cent on FrontierCode 1.1 Main, compared with 34.4 per cent for Gemini 3.6 Flash. On DeepSWE v1.1, it scored 65.3 per cent, compared with 49 per cent for the previous model.Also Read | Elon Musk’s answer to Claude Cowork: What is Grok Bot and what makes it different?The improvements seem to extend to web development too. Google said that Gemini 3.7 Flash can generate more functional layouts and feature-complete applications with fewer prompts. It can also closely reproduce the design of a reference image, screenshot or design system. On the WebDev Arena benchmark, it achieved an Elo score of 1,588, compared with 1,538 for Gemini 3.6 Flash.It needs to be noted that the model is not solely limited to coding. According to Google, it performs better on tasks involving areas such as finance, law and biosciences. On the GDP.pdf benchmark, which tests the ability to understand complex documents, Gemini 3.7 Flash scored 34 per cent, compared with 22 per cent for Gemini 3.6 Flash. On AutomationBench, which evaluates real-world business workflows, it scored 30.4 per cent, compared with 17 per cent for its predecessor.Why the focus on AI agents?Lately, AI agents are finding increasing use cases across a range of domains. These are systems that can do more than respond with crisp answers; they can instead plan and execute a chain of actions. Google said that the new Gemini 3.7 Flash has been designed to ‘think more diligently’ when dealing with tasks that require multi-step planning and tool calls. The model is claimed to be better at adapting when it encounters hurdles and also asks for clarifications when needed, following instructions with precision.These features make it particularly relevant for coding agents, enterprise automation and other applications where an AI system needs to complete a task rather than simply respond to a prompt. The search giant said that it is also using the model to power Gemini Spark, its personal AI agent that can work continuously under a user’s direction. Further, the company said that the updated model improves Spark’s ability to use Google Workspace tools, consolidate files, draft emails and update status documents.Story continues below this adCost and availabilityGemini 3.7 Flash comes at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. From January 1, 2027, those prices will increase to $1.50 per million input tokens and $7.50 per million output tokens. Gemini 3.7 Flash is half the original 3.6 Flash cost per million tokens. The previous model was announced only three weeks earlier, on July 21.Also Read | SpaceXAI unveils Grok 4.6, claims frontier-level performance at lower costHowever, Google’s argument is that developers can get stronger performance without paying the cost typically associated with more capable AI models. This could be vital for companies running AI agents at scale, where millions or billions of tokens can quickly translate into significant costs.Developers can access the model through Google’s Gemini API, Google AI Studio, Android Studio and Google Antigravity. Enterprises can access it through Google’s enterprise AI platforms, while individual users can use it through Gemini Spark with eligible Google AI Pro and Ultra subscriptions in supported markets. The company has also updated the model’s safeguards, including protections covering potential misuse involving chemical, biological, radiological and nuclear risks and cyber-related threats.In simple words, Gemini 3.7 Flash is less about being a general-purpose chatbot and more about becoming a high-performance engine for coding and AI agents.