Gemini 3.5 Flash (Vertex AI, US)

Gemini 3.5 Flash marks a generational leap within Google's Gemini Flash lineup, positioning the lightweight tier ahead of the previous Pro generation on agent-oriented evaluations. A Flash-tier Gemini reached 76.2% on Terminal-Bench 2.1 while generating output roughly four times faster than its predecessor, suggesting that reasoning depth and latency have both improved rather than trading off against each other. It defaults to medium thinking effort rather than high, and inherits the roughly one-million-token context window that has become standard across recent Gemini releases, making it well suited for long-document analysis and multi-step agent workflows where speed matters.

The model is designed for production use cases that need multimodal input handling, tool use, and structured outputs at a modest price point. It runs across thirty-seven provider routes on models.dev, with consistent support for reasoning, function calling, structured output, and temperature control, indicating a stable API surface that teams can integrate without specialized workarounds. Its Flash positioning—low cost, fast generation, broad capability set—makes it a practical fit for high-throughput assistants, retrieval-augmented pipelines, and code or agent tasks where the previous Pro generation was once the default.