Google Launches Gemini 3.5 Flash

Gemini 3.5 Flash: Google's Fastest Agentic Model Reviewed

Google DeepMind has released Gemini 3.5 Flash, the fastest agentic AI model in the Gemini family, benchmarked at four times the output token speed of competing frontier models and scoring 84.2% on CharXiv Reasoning. The model runs globally in the Gemini app, AI Mode in Search, and via the Gemini API in AI Studio for developers. API pricing starts at $0.15 per million input tokens and $1.25 per million output tokens. StartupHub.ai data scores Google DeepMind at 83 out of 100 in our AI lab rankings, with a 97 quality sub-score, the highest we record for any foundation model provider.

Gemini 3.5 Flash is positioned to deliver frontier-level performance for agents and coding tasks, handling intricate, long-horizon assignments that offer tangible real-world utility. The model is now accessible globally to consumers via the Gemini app and AI Mode in Google Search. Developers can access it through Google's Antigravity platform and the Gemini API in AI Studio and Android Studio, while enterprises can leverage it via Gemini Enterprise platforms.

Frontier Intelligence, Exceptional Speed

The new model boasts intelligence comparable to leading large language models but operates at the accelerated speeds characteristic of the Flash series. Google claims it is their most potent agentic and coding model to date, outperforming Gemini 3.1 Pro on key AI model benchmarks like Terminal-Bench 2.1 and GDPval-AA. Furthermore, it demonstrates strong multimodal understanding, achieving 84.2% on CharXiv Reasoning. Crucially, Gemini 3.5 Flash is reported to be four times faster than other frontier models in terms of output tokens per second, effectively eliminating the trade-off between quality and latency.

Agentic Tasks at Scale

This blend of speed and performance makes Gemini 3.5 Flash ideal for demanding agentic workflows. Tasks that previously took days or weeks can now be completed in a fraction of the time, often at reduced costs compared to other advanced models. It rapidly plans, builds, and iterates on solutions for real-world problems, from application development to financial document preparation.

When integrated with the updated Antigravity harness, Gemini 3.5 Flash can deploy collaborative subagents to tackle problems at scale. Under supervision, it reliably executes multi-step workflows and coding tasks while maintaining high performance. Examples include automatically renaming and categorizing unstructured assets, synthesizing research papers into playable games, and transforming legacy codebases.

Richer Graphics and Real-World Impact

Building on Gemini 3's multimodal capabilities, 3.5 Flash generates more interactive web UIs and graphics. This includes creating interactive animations, converting text descriptions into functional hardware designs, and rapidly developing branding concepts and UX approaches.

The model's agentic capabilities are already demonstrating significant value for enterprises. Partners like Shopify are using it for long-horizon data analysis to improve merchant growth forecasts. Macquarie Bank is piloting its use for accelerated customer onboarding by processing complex documents. Salesforce is integrating Gemini 3.5 Flash into its Agentforce to automate enterprise tasks with multi-agent coordination, while Ramp is enhancing OCR for invoices, and Xero is automating tax form preparation for small businesses. Databricks is employing agentic workflows for real-time data monitoring and issue diagnosis.

Personal AI Agents and Safety

Gemini 3.5 Flash now serves as the default model for the Gemini app and AI Mode in Search globally. New features, such as the personal AI agent Gemini Spark, leverage its capabilities to assist users with daily digital tasks. Google is also enhancing generative UI experiences in Search with its coding prowess.

Developed under Google's Frontier Safety Framework, Gemini 3.5 incorporates strengthened cyber and CBRN safeguards. Advanced safety training and mitigation techniques, including interpretability tools, aim to reduce harmful content generation and prevent unwarranted refusals of safe queries.

Gemini Flash After 3.5: What Came Next

Gemini 3.5 Flash was followed by two successor models in quick succession. Google released Gemini 3.6 Flash on July 21, 2026, targeting coding, tool use, long context, and mixed media workflows. It supports 1,048,576 input tokens and outputs up to 65,536 tokens per call, with API pricing at $1.50 per million input tokens and $7.50 per million output tokens. Three weeks later, on August 13, 2026, Google released Gemini 3.7 Flash at roughly half the cost of 3.6 Flash, citing significant coding benchmark improvements. For developers starting a new project today, 3.7 Flash is the current recommended default in Google's Gemini API changelog. Gemini 3.5 Flash remains available as a stable API model and continues to be used in production by partners who built against its API ID before the 3.6 and 3.7 releases.

Frequently Asked Questions

What is Gemini 3.5 Flash?

Gemini 3.5 Flash is Google DeepMind's agentic AI model designed for high-speed, complex task execution. It runs at four times the output token throughput of competing frontier models, scores 84.2% on CharXiv Reasoning, and outperforms Gemini 3.1 Pro on Terminal-Bench 2.1 and GDPval-AA benchmarks. It is available via the Gemini API in AI Studio and Android Studio, and as the default model in the Gemini consumer app globally.

How much does the Gemini 3.5 Flash API cost?

Gemini 3.5 Flash API pricing starts at $0.15 per million input tokens and $1.25 per million output tokens, making it one of the more cost-efficient frontier-grade models for agentic workloads. Developers can access it via Google AI Studio for free within quota limits before committing to paid API usage.

How does Gemini 3.5 Flash compare to Gemini 3.6 and 3.7 Flash?

Gemini 3.6 Flash (released July 21, 2026) added longer context windows (up to 1,048,576 input tokens) and stronger coding and tool-use performance at $1.50 per million input tokens. Gemini 3.7 Flash (released August 13, 2026) cut the cost roughly in half versus 3.6 while improving coding benchmarks by 16 points on real software engineering evaluations. For new projects, 3.7 Flash is the current recommended model. Gemini 3.5 Flash remains a stable, production-proven option for teams already integrated with its API.

What are the best agentic use cases for Gemini 3.5 Flash?

Gemini 3.5 Flash is built for multi-step agentic workflows where latency matters: coding tasks, document processing, financial analysis, real-time data monitoring, and long-horizon planning assignments. Enterprise partners including Shopify (merchant growth forecasting), Macquarie Bank (customer onboarding), and Salesforce (Agentforce automation) have deployed it in production. Its speed advantage over Pro-tier models makes it the preferred choice for agent loops that make many sequential model calls.