Gemini 3.7 Flash Specs and Pricing

Google shipped Gemini 3.7 Flash on August 13, 2026, three weeks after 3.6 Flash and while Gemini 3.5 Pro remains delayed. The company calls it “our most intelligent workhorse model,” and the official announcement backs the claim with benchmark jumps of 10 to 16 points on coding and agent tasks. If you build against the Gemini API, this is the model you’ll be pinning for the next few months.

This page is the reference card. Every spec, price, benchmark delta, and access channel in one place, in tables you can scan in under a minute. When you’re ready to write code, the Gemini 3.7 Flash API quickstart walks through cURL, Python, and Node step by step, and Apidog gives you a workspace to fire test requests against the endpoint before any of that code exists.

Bookmark this one. The numbers below change on January 1, 2027, when the introductory pricing window closes, and we’ll keep the tables current.

TL;DR

• Model ID is gemini-3.7-flash, live in the Gemini API as of August 13, 2026, available in 160+ countries.
• Specs: 1M token input context, 64k token output limit, multimodal input (text, image, video, audio, PDF), text output.
• Tools: function calling, search as a tool, and computer use are all supported.
• Introductory pricing: $0.75 per 1M input tokens and $3.75 per 1M output tokens through December 31, 2026. Standard rates double to $1.50 and $7.50 on January 1, 2027.
• Benchmarks over 3.6 Flash: DeepSWE v1.1 49.0% to 65.3%, AutomationBench 17.0% to 30.4%, WebDev Arena Elo 1538 to 1588.
• Access it through the Gemini API, Google AI Studio, the Gemini app, Gemini Spark, Google Antigravity, Android Studio, and Gemini Enterprise.

Gemini 3.7 Flash spec sheet

The core numbers, straight from Google’s Gemini Flash model page and the launch materials.

Two rows deserve a second look. The 64k output limit is generous for a Flash-class model; it means full file rewrites and long structured outputs fit in a single response. And computer use support puts 3.7 Flash in a small group of models that can drive a browser directly, though for most production automation a structured API call still wins. The tradeoffs are covered in computer use vs structured APIs.

Pricing at a glance

Google launched 3.7 Flash at half the launch price of 3.6 Flash, but the discount has an expiry date.

That’s a clean 2x jump on both sides on New Year’s Day. If your workload is token-heavy, the intro window is the time to benchmark, tune prompts, and lock in caching strategies while mistakes cost half as much. For worked cost examples, token math on real workloads, and strategies to cut spend before the standard rates hit, see Gemini 3.7 Flash pricing explained. The live source of truth is the Gemini API pricing page.

Benchmarks: 3.6 Flash vs 3.7 Flash

Google published deltas against 3.6 Flash across coding, web development, document understanding, and agentic automation. The gaps are large for a point release that landed three weeks after its predecessor, a point 9to5Google’s launch coverage also flagged.

The pattern across the table: agent workloads gained the most. Google’s stated improvements match, better debugging, more deployable first-try code, more diligent thinking on multi-step planning and tool calls, and tighter instruction following. How those numbers stack up against Anthropic and OpenAI models is a separate question, answered in Gemini 3.7 Flash vs Claude vs GPT.

Where you can use Gemini 3.7 Flash

Seven access channels at launch. Pick by role.

For API work, the AI Studio key path gets you a working request in minutes. Vertex AI carries the same model behind OAuth with IAM, audit logs, and regional endpoints when you need production controls.

A minimal working request

Grab a key from AI Studio, export it, and send one request.

export GEMINI_API_KEY="AIza..."

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.7-flash:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "parts": [{ "text": "List the HTTP status codes an idempotent retry layer should handle." }]
    }],
    "generationConfig": {
      "temperature": 0.4,
      "maxOutputTokens": 1024
    }
  }'

Streaming swaps the method for :streamGenerateContent?alt=sse on the same path. The response returns a candidates array with the text parts and a usageMetadata block with the token counts you’ll be billed on.

If you’d rather not hand-edit JSON in a terminal, Apidog imports the Generative Language API spec, binds GEMINI_API_KEY as an environment variable on the x-goog-api-key header, and renders SSE streams live. Save a good response as an example once and rerun tests against the fixture instead of the paid endpoint.

Rate limits and tiers

Google sets rate limits per model and per tier (free, pay-as-you-go, and provisioned levels), and the numbers shift often enough that any table printed here would rot. Check the official Gemini API docs for the current requests-per-minute and tokens-per-minute caps on gemini-3.7-flash.

Two stable rules of thumb. Free-tier quota through AI Studio is enough to prototype but not to ship. And a 429 with status RESOURCE_EXHAUSTED means you’ve hit a per-minute or per-day cap; back off with jitter, batch where you can, or move the project to billed tier.

Quick links

• Official announcement: launch post with Google’s claims and benchmark charts
• Gemini API docs: endpoint reference, SDK guides, rate limits
• Pricing page: live per-token rates
• Model page: specs and modality details
• Get an API key: AI Studio key in two clicks

FAQ

What are the Gemini 3.7 Flash specs in one sentence?

A 1M token input context window, a 64k token output limit, multimodal input across text, image, video, audio, and PDF, text output, and support for function calling, search as a tool, and computer use, addressable as gemini-3.7-flash through the Gemini API.

How much does Gemini 3.7 Flash cost?

$0.75 per 1M input tokens and $3.75 per 1M output tokens through December 31, 2026. From January 1, 2027 the standard rates apply: $1.50 input and $7.50 output. The intro rate is half of what 3.6 Flash launched at.

Do I need to change my code to upgrade from 3.6 Flash?

The swap itself is one line: change the model ID to gemini-3.7-flash. The behavior changes are the real work, since 3.7 thinks longer on multi-step tool calls and follows instructions more strictly. The 3.6 to 3.7 Flash migration guide covers what to regression-test before you cut traffic over.

Why did Google ship 3.7 Flash before Gemini 3.5 Pro?

Gemini 3.5 Pro is still delayed, and Axios reported that Google is shipping Flash updates ahead of its next flagship rather than holding improvements back. The three-week gap between 3.6 and 3.7 Flash fits that cadence.

Is there a free way to try it?

Yes. AI Studio’s free tier includes daily quota for gemini-3.7-flash, enough for prompt testing and small prototypes, and no billing account is required until you need higher rate limits. The free Gemini API access guide covers how far the free tier stretches and what to do when you hit its caps.

Where 3.7 Flash fits in your stack

Treat 3.7 Flash as the default Gemini model for agent loops, coding assistants, document pipelines, and anything that makes many tool calls per session. The benchmark gains land exactly where those workloads live, and the intro pricing makes the next four and a half months the cheapest window to build in. Reserve judgment on flagship-class reasoning until Gemini 3.5 Pro ships; for everything else, this is the workhorse Google says it is.

Before the model ID goes anywhere near production code, put the endpoint under test. Download Apidog to keep your Gemini requests, environments, streaming checks, and saved response fixtures in one workspace, alongside collections for every other provider you’re evaluating.