Gemini 3.7 Flash

Why teams choose it

Tradeoffs to know

• Google's launch benchmark comparisons are vendor-reported and should be reproduced on representative workloads.
• Availability, quotas, safety controls, and feature parity can differ across Gemini API, enterprise, and consumer surfaces.
• The introductory price changes in 2027; model total cost also depends on reasoning, tools, caching, and retries.

When not to use this

• Not ideal for sprawling research or brittle multi-hop reasoning unless you constrain scope tightly.
• Avoid for regulated or high-stakes outputs without evaluations that mimic your tooling, data, and review process.
• Promote traffic to heavier tiers inside the family when workflows need richer tools and longer horizons.

Technical specs

Inputs

text, image, video, audio, PDF

Outputs

text

Capabilities

coding, agentic workflows, web development, knowledge work, multimodal reasoning, tool use, document analysis, structured output

License

Proprietary API

Model string

gemini-3-7-flash

Benchmarks

{
  "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/",
  "gdp_pdf": 34,
  "deep_swe_v1_1": 65.3,
  "vendor_reported": true,
  "webdev_arena_elo": 1588,
  "frontier_code_1_1_main": 43.6
}