Gemini 3.6 Flash

gemini-3.6-flash

Gemini 3.6 Flash continues to deliver frontier model-level intelligence, optimized to handle real-world tasks at faster speeds and lower costs. Designed for the agent era, it excels at code generation, agent execution, and spatial reasoning. This model is particularly effective in fast agent loops involving complex coding cycles and iterations.

Context window105K

ProviderGemini

Released2026/08/10

Playground

Pricing

The same model is available through multiple service channels — choose based on latency, reliability and cost.

Prices in $ / 1M tokens

To pick a channel, add a provider field to the request body, for example "provider": { "channel": "direct" }. Valid values are direct / stable / economical; omit it to use the default channel.

Direct

Direct upstream connection — best when you need native behavior and the full context window.

Input	Output	Cache read
1.50/M	7.50/M	0.15/M

Overview

Input

Text imagve

Output

Text

Gemini 3.6 Flash API: The Token-Saving Workhorse Model for the Agent Era

Gemini 3.6 Flash is an efficient multimodal model released by Google on July 21, 2026, positioned as the workhorse tier "built for the agent era." It surpasses Gemini 3.5 Flash across coding, knowledge work, and multimodal understanding, while reducing output token usage by approximately 17% — for the same tasks, it answers better and costs less.

It is specifically optimized for multi-step orchestration, full-stack code refactoring, and general reasoning, and offers Computer Use as a built-in client tool. It supports text, image, video, audio, and PDF inputs, with a 1 million token context window.

SeaWhale AI provides Gemini 3.6 Flash through an OpenAI-compatible interface, supporting tool calling, streaming output, and multimodal input.

Get API Key · Model ID: gemini-3.6-flash

Why Choose Gemini 3.6 Flash

17% fewer output tokens — Completes tasks with fewer reasoning steps than 3.5 Flash, directly lowering your bill
Significantly better code quality — 49% on DeepSWE code quality (vs. 37% for 3.5 Flash), with lower compilation failure and rework rates
MLE Bench 63.9% — A jump from 49.7% on machine learning engineering tasks
OSWorld-Verified 83.0% — Computer use capability in real operating system environments (78.4% for 3.5 Flash)
GDPval-AA v2 1421 Elo — Real-world professional knowledge work performance (1349 for 3.5 Flash)
Full-modality input + 1M context — Unified processing of text, images, video, audio, and PDF; up to 64K output tokens per request

Core Capabilities

01 Coding and Full-Stack Refactoring

Gemini 3.6 Flash's improvement in code generation isn't just "getting it right" — it's "getting it right the first time": lower compilation failure rates, fewer revision rounds, and improvements of up to 65% in some DeepSWE scenarios.

Multi-file code generation and full-stack refactoring
Lower compilation failure and rework rates
Machine learning engineering tasks (MLE Bench 63.9%)

02 Agent Execution and Computer Use

Computer Use is available directly as a built-in client tool, with OSWorld-Verified 83.0% reflecting its reliability in clicking, typing, and navigating in real desktop environments.

Built-in Computer Use tool
Multi-step task orchestration with fewer reasoning steps
Full support for tool calling, structured output, and code execution

03 Full-Modality Understanding

Supports text, image, video, audio, and PDF inputs, and is significantly stronger than the previous generation on multimodal reasoning tasks such as chart interpretation, converting visual blueprints to code, and generating multi-element web layouts.

Chart and document parsing
Visual design mockup to frontend code
Video and audio content understanding

04 Adjustable Thinking Modes

Fully supports adjustable thinking modes, allowing precise allocation of compute based on task difficulty; combined with its inherent higher token efficiency, the cost advantage is further amplified in high-frequency calling scenarios.

Increase thinking depth for difficult tasks
Lower the mode for latency-sensitive tasks to speed up responses
Full feature set including search grounding, URL context, and caching

Best Use Cases

Scenario	Description
High-throughput coding assistant	Gets it right the first time, less rework, fewer tokens
Agent automation	Multi-step orchestration + built-in computer use tool
Knowledge workflows	Reports, analysis, document processing (GDPval-AA v2 1421)
Chart and document parsing	Visual reasoning on PDFs, charts, and design mockups
Machine learning engineering	Data processing, modeling, experiment scripts (MLE Bench 63.9%)
Cost-sensitive production deployment	The 17% output token reduction shows up directly on your bill

Gemini 3.6 Flash vs. Gemini 3.5 Flash

Capability	Gemini 3.6 Flash	Gemini 3.5 Flash
Model ID	`gemini-3.6-flash`	`gemini-3.5-flash`
Positioning	Workhorse tier for the agent era	Efficient multimodal workhorse
Output token efficiency	~17% reduction	Baseline
DeepSWE code quality	49%	37%
MLE Bench	63.9%	49.7%
OSWorld-Verified	83.0%	78.4%
GDPval-AA v2	1421 Elo	1349 Elo
Computer use	Built-in client tool	—
Input modalities	Text, image, video, audio, PDF	Text, image, video, audio, PDF
Context window	1M tokens / 64K output	1M tokens

Refer to the real-time price card at the top of the page for specific billing details.

How to Use the Gemini 3.6 Flash API

1. Create a SeaWhale AI API key Generate a key in the console and add credits.

2. Start with the default thinking mode The default mode is sufficient for most tasks, and combined with higher token efficiency it's usually cheaper than 3.5 Flash; only raise the thinking depth when quality falls short.

3. Call the API

curl -X POST https://api.atalk-ai.com/v1/chat/completions \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "gemini-3.6-flash",
    "messages": [
      {"role": "user", "content": "Break down this requirements document into a frontend and backend task list, and provide the API definition for each task."}
    ],
    "stream": true
  }'

FAQ

When was Gemini 3.6 Flash released? July 21, 2026, released alongside Gemini 3.5 Flash-Lite and 3.5 Flash Cyber.

How is it better than 3.5 Flash? Three main areas: code quality (DeepSWE 49% vs. 37%, with up to 65% improvement in some scenarios), knowledge work (GDPval-AA v2 1421 vs. 1349), and token efficiency (~17% output reduction). Computer use has also evolved from an external add-on to a built-in tool.

How large is the context window? 1 million input tokens (1,048,576), with up to 65,536 output tokens per request.

Which input modalities are supported? Text, image, video, audio, and PDF; output is text. Image/audio generation is not supported.

Which developer features are supported? Function calling, structured output, context caching, code execution, computer use (preview), search grounding, URL context, batch API, and more.

Is it suitable for building agents? Very much so. It was designed for agent scenarios: multi-step orchestration with fewer reasoning steps, and OSWorld-Verified 83.0% computer use capability can directly drive desktop automation.

Why Use Gemini 3.6 Flash API on SeaWhale AI

Direct connection in China — No Google Cloud account or overseas network required
OpenAI-compatible interface — Integrate by changing just two lines in your existing code
Unified entry point for all modalities — Video, audio, PDF, and text share a single API
Cross-vendor comparison — Compare pricing and performance side-by-side with GPT and Claude series under the same account

API

API integration

Model IDUse this value as the model in inference requests

gemini-3.6-flash

API KeyBearer token used to authenticate inference requests

Manage API keys

Base URLOpenAI compatible · /chat/completions

https://api.atalk-ai.com/v2/

provider OptionalSelects a service channel; omit it and the system picks the default

"provider": { "channel": "direct" }

gemini-3.6-flash usage examples

SeaWhale AI is compatible with the OpenAI API protocol, so you can call it with the OpenAI SDK or plain HTTP requests. Streaming is enabled by default.

About the provider parameter (optional, a SeaWhale AI extension): most models are served over several channels that differ slightly in price and reliability. Add a provider field to the request body to pick one; omit it and the system selects the default channel — normal calls are unaffected.

provider is not part of the official OpenAI protocol — it is a SeaWhale AI extension that only takes effect on this platform. The OpenAI SDK allows custom fields like this to pass through; see the examples below.

Value	Channel	Best for
`direct`	Direct	The official upstream link, for native behavior and the full context window
`stable`	Preferred	Balanced availability and speed — a good default for production traffic
`economical`	Economy	Cost first, well suited to batch processing and price-sensitive workloads

Available channels and their prices are listed under "Pricing" above (channels vary by model). Additional notes:

Syntax: "provider": { "channel": "direct" }.
If the requested channel is not enabled for that model, the request automatically falls back to the default channel and returns normally — it does not error.
With the official SDKs: in Python pass it via extra_body; in Node.js put it directly on the request object and it passes through. In TypeScript projects, add a // @ts-expect-error line to skip the type check.

cURLPythonNode.js

curl https://api.atalk-ai.com/v2/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <API_KEY>" \
  -d '{
    "model": "gemini-3.6-flash",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ],
    "provider": { "channel": "direct" },
    "stream": true
  }'
# provider is optional — remove this line to use the default channel

from openai import OpenAI

client = OpenAI(
    base_url="https://api.atalk-ai.com/v2",
    api_key="<API_KEY>",
)

stream = client.chat.completions.create(
    model="gemini-3.6-flash",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello!"},
    ],
    stream=True,
    # Optional: pick a service channel; omit to use the default
    extra_body={"provider": {"channel": "direct"}},
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://api.atalk-ai.com/v2',
  apiKey: '<API_KEY>',
})

const stream = await client.chat.completions.create({
  model: 'gemini-3.6-flash',
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'Hello!' },
  ],
  stream: true,
  // Optional: pick a service channel; omit to use the default
  // @ts-expect-error provider is a SeaWhale AI extension, not in the OpenAI SDK types
  provider: { channel: 'direct' },
})

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? '')
}