Google DeepMind Introduces Gemini 3.7 Flash, Its Most Intelligent Workhorse Model for Coding and Agents

Google DeepMind has announced the release of Gemini 3.7 Flash, described by the company as its most intelligent workhorse model to date for coding and agent-driven workflows. The announcement was made on August 13, 2026, by Tulsee Doshi, Senior Director of Product Management, on behalf of the Gemini team.

The launch arrives just three weeks after the release of Gemini 3.6 Flash, continuing a cadence of near-monthly Flash-tier updates from DeepMind. The compressed interval between releases reflects an industry pattern in which frontier model providers are shipping incremental but measurable improvements on faster cycles, often motivated by developer feedback rather than multi-month research milestones. According to Google DeepMind, 3.7 Flash is a direct result of developer feedback and algorithmic innovations that the company plans to carry forward into future models. The new release delivers substantial improvements across software engineering, knowledge work, and web development workflows, with an introductory price set at half the original 3.6 Flash cost per million tokens.

Performance Gains in Coding, Web Development, and Knowledge Work

Gemini 3.7 Flash demonstrates strong gains over its predecessor in coding tasks such as debugging and issue resolution. The model achieves higher first-pass code accuracy and improved performance in generating production-ready code, as measured by the FrontierCode 1.1 Main benchmark, where it scored 43.6% compared to 3.6 Flash's 34.4%. On DeepSWE v1.1 — an evaluation that tests a model's ability to resolve real GitHub issues end-to-end — 3.7 Flash scored 65.3% versus 3.6 Flash's 49.0%.

In web development, 3.7 Flash generates more functional layouts and feature-complete applications in fewer prompts. For UI generation, the model exhibits high design adherence and parity relative to reference inputs, including screenshots, images, and full design systems. On Arena.ai's WebDev Arena, 3.7 Flash achieved an Elo score of 1588, compared to 1538 for 3.6 Flash.

For knowledge-dense fields such as finance, law, and biosciences, 3.7 Flash delivers improved reasoning and accuracy. It significantly outperforms 3.6 Flash on the GDP.pdf benchmark — an evaluation designed to test a model's ability to process complex documents — scoring 34.0% versus 22.0%. The model also surpasses 3.6 Flash on AutomationBench, which measures the ability to complete real-world business workflows, scoring 30.4% compared to 17.0%.

Google DeepMind highlighted several use cases enabled by 3.7 Flash, including generating fully playable 3D games from text prompts when combined with Nano Banana, creating interactive landing pages in a single shot by orchestrating sub-agents with Gemini Omni, training robotics models using multimodal understanding in a three-agent graph loop, and transforming static PDFs such as annual reports into interactive web experiences with live charts and aggregated insights.

Developer Experience and Pricing

Gemini 3.7 Flash offers a noticeably improved developer experience over 3.6 Flash, according to the company. The model better adapts to roadblocks, clarifies intent when needed, and follows instructions with greater fidelity. It also demonstrates more diligent multi-step planning and tool-call execution, resulting in less manual oversight and fewer retries across engineering workflows.

3.7 Flash is available through the end of 2026 at an introductory price of $0.75 per 1 million input tokens and $3.75 per 1 million output tokens. Google DeepMind stated that this pricing, combined with the model's enhanced performance, enables developers and customers to scale production-ready agents cost-effectively. Early customer feedback has reportedly highlighted 3.7 Flash's performance and precision, achieving results significantly better than 3.6 Flash at a low cost.

Introductory pricing expires on December 31, 2026. Starting January 1, 2027, the rate will be $1.50 per 1 million input tokens and $7.50 per 1 million output tokens.

Gemini Spark Integration

Gemini Spark, available to Google AI Pro and Ultra subscribers in over 160 countries, began using Gemini 3.7 Flash on August 13, 2026. Originally launched at Google I/O, Spark is described as a personal AI agent that operates 24/7, taking action on a user's behalf under their direction. The model update is designed to make Spark more efficient for knowledge work, with improved tool use for Google Workspace applications, delivering enhanced accuracy and output quality for complex, multi-skill workflows. With 3.7 Flash, Spark can consolidate files, draft emails, and update status documents more efficiently.

Safety Measures

Google DeepMind stated that it continually works to improve the coverage and robustness of its Frontier Safety safeguards. Gemini 3.7 Flash ships with updated safeguards against misuse in the domains of Chemical, Biological, Radiological, and Nuclear (CBRN) threats and cyber offense, while enabling beneficial use cases, in accordance with the company's approach to bioresilience and its cyber program. Additional details are available in the 3.7 Flash model card.

Availability

• Developers: Explore agent-first workflows in Google Antigravity or start building via the Gemini API through Google AI Studio and Android Studio. A developer guide is available.
• Enterprises: Access 3.7 Flash through the Gemini Enterprise Agent Platform and the Gemini Enterprise app.
• Individuals: Available via Spark in the Gemini app for Google AI Pro and Ultra subscribers in supported countries.

Source: Google DeepMind Blog