DEV Community

Cover image for What's New in Gemini 3.7 Flash? Features, Benchmarks, and API Access
Hassann
Hassann

Posted on Originally published at apidog.com

What's New in Gemini 3.7 Flash? Features, Benchmarks, and API Access

Gemini 3.7 Flash is Google’s newest workhorse AI model, released on August 13, 2026, three weeks after Gemini 3.6 Flash. It keeps the 1M token context window and multimodal input of its predecessor while posting large gains on coding and agent benchmarks, and it launches at an introductory API price of $0.75 per 1M input tokens. Google calls it “our most intelligent workhorse model.”

Try Apidog today

That three-week gap between releases is the story. Google is shipping Flash updates on a sprint cadence while Gemini 3.5 Pro stays delayed, which means the mid-tier model is now where the interesting engineering lands first. The benchmark deltas back that up: DeepSWE jumps 16 points, AutomationBench nearly doubles, and the intro pricing undercuts the 3.6 Flash launch rate by half.

This post covers what changed, the benchmark results, the pricing windows to calendar, where to access the model, and a cURL request you can run in five minutes. For a deeper endpoint walkthrough, see the Gemini 3.7 Flash API quickstart. You can also use Apidog to test the new model against your current prompts before switching production traffic.

TL;DR

  • Gemini 3.7 Flash launched August 13, 2026, three weeks after 3.6 Flash. Model ID: gemini-3.7-flash.
  • Coding and agent benchmarks improved: DeepSWE v1.1 went from 49.0% to 65.3%, AutomationBench from 17.0% to 30.4%, and WebDev Arena from 1538 to 1588 Elo.
  • Intro API pricing is $0.75 per 1M input tokens and $3.75 per 1M output tokens—half of 3.6 Flash’s launch price. Standard rates ($1.50 / $7.50) start January 1, 2027.
  • Specs are unchanged: 1M-token input context, 64k output limit, multimodal input, function calling, search as a tool, and computer use.
  • It is available in the Gemini API, AI Studio, Gemini app, Gemini Spark, Google Antigravity, Android Studio, and Gemini Enterprise across 160+ countries.
  • Gemini 3.5 Pro is still delayed, while Google continues shipping Flash improvements.

What Gemini 3.7 Flash is

Flash is the middle tier of the Gemini line: cheaper and faster than Pro, smarter than Flash-Lite, and tuned for high-volume production workloads. Gemini 3.7 Flash is the third Flash release in the 3.x line.

The official announcement positions it as a coding and agent model first, and a chat model second.

Gemini 3.7 Flash overview

The spec sheet carries over from 3.6 Flash:

  • Context: 1M-token input window and 64k-token output limit.
  • Input modalities: text, image, video, audio, and PDF. Output is text.
  • Tooling: function calling, search as a tool, and computer use.
  • Safety: updated CBRN and cyber safeguards ship with the release.

What changed is behavior rather than architecture. Google says 3.7 Flash debugs better, produces deployable code on the first try more often, adapts when it hits roadblocks, asks clarifying questions for ambiguous intent, and follows instructions more reliably.

The benchmarks are the part you can validate with your own prompt suite.

Benchmark improvements: 3.6 vs. 3.7

Google spread these results across a blog post and model card. Here are the key deltas:

Benchmark Gemini 3.6 Flash Gemini 3.7 Flash Change
DeepSWE v1.1 (agentic coding) 49.0% 65.3% +16.3 pts
FrontierCode 1.1 Main 34.4% 43.6% +9.2 pts
WebDev Arena (Elo) 1538 1588 +50 Elo
GDP.pdf (document reasoning) 22.0% 34.0% +12.0 pts
AutomationBench (agent tasks) 17.0% 30.4% +13.4 pts

Two standalone scores add context:

  • 90.7% on Harvey LAB-AA, a legal reasoning benchmark.
  • 97.0% on 128k-needle long-context retrieval.

The largest gains are on multi-step agent tasks. AutomationBench moving from 17.0% to 30.4% may change which workflows are viable for a mid-tier model. Launch coverage highlighted the DeepSWE result for the same reason: a 16.3-point gain in agentic coding over three weeks is notable.

For cross-model benchmark context, see Grok 4.6 vs GPT 5.6 vs Claude Fable 5.

Coding and agent improvements

Use the benchmark changes to define a targeted evaluation instead of relying on generic chat prompts.

Test these areas:

  • Debugging: Give the model failing tests and multi-file repositories. DeepSWE is the closest benchmark signal because it measures multi-file bug fixes.
  • First-try deployable code: Ask for implementation changes with explicit acceptance criteria, then run the generated code. FrontierCode improved by 9.2 points.
  • Roadblock handling: Simulate failed tool calls, missing files, or permission errors. Check whether the model changes its plan instead of repeating the same failed step.
  • Intent clarification: Send ambiguous requirements. Measure whether it asks a useful question before making assumptions.
  • Instruction fidelity: Test output schemas, length limits, negative constraints, and formatting requirements across longer conversations.

The tool stack is also unchanged from 3.6 Flash:

  • Function calling for structured integrations.
  • Search as a tool for retrieval workflows.
  • Computer use for browser-driven tasks when an API does not exist.

Computer use can be useful, but it is slower and less deterministic than a structured API integration. See computer use vs. structured APIs to decide which approach fits a workflow.

Pricing: half price until December 31, 2026

Gemini 3.7 Flash has a two-phase Gemini API pricing schedule:

Period Input per 1M tokens Output per 1M tokens
Through December 31, 2026 $0.75 $3.75
From January 1, 2027 $1.50 $7.50

The introductory price is half of Gemini 3.6 Flash’s launch price. However, the cost doubles on January 1, 2027.

Before deploying:

  1. Verify current pricing on the official Gemini API pricing page.
  2. Forecast long-lived workloads using the standard 2027 rate.
  3. Record input and output tokens for each request from API usage metadata.
  4. Compare cost per successful task, not only cost per token.

For worked token-cost examples, see the Gemini 3.7 Flash pricing breakdown.

Where you can use it today

Google shipped 3.7 Flash across its product surface area in 160+ countries:

  • Gemini API: Get a key from AI Studio and call gemini-3.7-flash.
  • Google AI Studio: Browser playground for prompt testing.
  • Gemini app: Consumer chat app access.
  • Gemini Spark: Available to AI Pro and Ultra subscribers.
  • Google Antigravity: Google’s agentic development environment.
  • Android Studio: IDE code assistance.
  • Gemini Enterprise: Managed organizational offering.

For API access at scale, Vertex AI provides a path through aiplatform.googleapis.com with OAuth, IAM, and audit logging. The model and request schema are the same, but authentication and hosting guarantees differ.

Why Google shipped Flash before Gemini 3.5 Pro

The release order is unusual: 3.6 Flash arrived in late July, 3.7 Flash followed three weeks later, and Gemini 3.5 Pro still has no date.

Axios reports that Google is shipping Flash updates while its next flagship remains delayed. For developers, the practical takeaway is straightforward: do not treat Flash as the trailing edge of the model lineup. Agent capabilities are landing there first.

If you deferred a migration from 3.6 Flash while waiting for Pro, start with the 3.6 to 3.7 Flash migration guide. For most codebases, the migration is a model ID change followed by regression testing.

How to try the API in five minutes

Create an API key in AI Studio, export it, then send a 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": "Review this function for bugs: def dedupe(items): return list(set(items))"
      }]
    }],
    "generationConfig": {
      "temperature": 0.4,
      "maxOutputTokens": 1024
    }
  }'
Enter fullscreen mode Exit fullscreen mode

The response includes:

  • A candidates array containing generated content in content.parts.
  • A usageMetadata object with exact input and output token counts.

Capture usageMetadata from day one. It is your cost meter.

To stream output, replace :generateContent with :streamGenerateContent?alt=sse:

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.7-flash:streamGenerateContent?alt=sse" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "parts": [{ "text": "Explain server-sent events in three bullet points." }]
    }]
  }'
Enter fullscreen mode Exit fullscreen mode

Once the request works, move comparisons out of the terminal. In Apidog:

  1. Import the Generative Language API specification.
  2. Store GEMINI_API_KEY as an environment variable.
  3. Bind it to the x-goog-api-key header.
  4. Save the model ID as a path variable.
  5. Run identical requests against gemini-3.6-flash and gemini-3.7-flash.
  6. Save outputs as examples for regression testing.

This gives you a direct comparison of output quality, instruction following, streaming behavior, latency, and token usage using your own workload.

FAQ

Is Gemini 3.7 Flash free to use?

The Gemini API has a free tier through AI Studio with daily quota for prototyping. Paid usage starts at $0.75 per 1M input tokens through December 31, 2026.

For guidance on working within quota limits, see free Gemini API access.

What’s the difference between Gemini 3.6 Flash and 3.7 Flash?

The model specs are the same: context window, output limit, modalities, and tool support remain unchanged.

The differences are benchmark and behavior improvements:

  • DeepSWE: +16.3 points
  • AutomationBench: +13.4 points
  • WebDev Arena: +50 Elo

Google also claims improved debugging, instruction following, and multi-step planning.

Does Gemini 3.7 Flash replace Gemini 3.5 Pro?

No. Pro remains the flagship tier for the hardest reasoning tasks, and Gemini 3.5 Pro is still in development.

Gemini 3.7 Flash targets high-volume production workloads where cost and latency matter alongside quality.

What happens to pricing on January 1, 2027?

The introductory price ends. Standard pricing becomes:

  • Input: $1.50 per 1M tokens
  • Output: $7.50 per 1M tokens

Budget at the standard rate for workloads that will continue beyond 2026.

Can Gemini 3.7 Flash process images and PDFs?

Yes. It accepts text, image, video, audio, and PDF inputs in the same contents array. Output is text only.

Google’s GDP.pdf result increased from 22.0% to 34.0%, which it presents as evidence of improved document reasoning.

Where 3.7 Flash fits in your stack

Gemini 3.7 Flash combines agent-focused benchmark improvements with unchanged integration specs and a temporary pricing discount.

The implementation path is simple:

  1. Run your existing prompt suite against gemini-3.7-flash.
  2. Compare outputs with your current model.
  3. Track token usage and latency.
  4. Test structured output, tool failures, long-context retrieval, and ambiguous instructions.
  5. Make the migration decision from measured results.

Use Apidog to save 3.6 Flash responses as examples, replay the same requests against 3.7 Flash, and validate whether the benchmark improvements hold for your workload.

Top comments (0)