Gemini 3.6 Flash and 3.5 Flash-Lite launch

Published: July 22, 2026
Last verified: July 23, 2026
Status: Superseded by the Gemini 3.8 Flash launch on September 3, 2026

Gemini 3.6 Flash and Gemini 3.5 Flash-Lite launched on LaoZhang API on July 22, 2026. At launch, this announcement recommended 3.6 Flash for general multimodal, coding, and multi-step agent work, and 3.5 Flash-Lite for classification, extraction, document parsing, and other high-volume work where cost or latency matters most.

What is the difference between the two models?

Model	Official status	Default thinking	Core position	Recommended workloads
Gemini 3.6 Flash (`gemini-3.6-flash`)	GA and production-ready	`medium`	Main balance of speed, intelligence, and token efficiency	Coding, multimodal understanding, complex agents, knowledge work
Gemini 3.5 Flash-Lite (`gemini-3.5-flash-lite`)	GA and production-ready	`minimal`	Lower-latency, lower-cost, high-throughput model	Subagents, classification, extraction, document parsing, bulk automation

Google documents an input context of about one million tokens and up to 64K output tokens for both models. Effective capabilities, concurrency, and billing on LaoZhang API still depend on the token group and current route configuration. Gemini 3.8 Flash is the newer Flash model. Both models on this page are still listed in public pricing as of September 24, 2026; see the model and pricing catalog.

Correct model IDs

Gemini 3.6 Flash: gemini-3.6-flash
Gemini 3.5 Flash-Lite: gemini-3.5-flash-lite

Use the complete model ID. Do not invent -preview, -thinking, or -nothinking suffixes; use a compatibility alias only when the console explicitly lists it.

How do I call the models through LaoZhang API?

LaoZhang API provides an OpenAI-compatible endpoint. This minimal request uses Gemini 3.6 Flash; replace only the model value to test Flash-Lite. Replace the placeholder with your LaoZhang API key:

export LAOZHANG_API_KEY="YOUR_LAOZHANG_API_KEY"

curl https://api.laozhang.ai/v1/chat/completions \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3.6-flash",
    "messages": [
      {"role": "user", "content": "Compare these two product requirements and return a risk checklist."}
    ]
  }'

The reply text is in choices[0].message.content. Before production rollout:

Confirm that the model is enabled for your token group on the console model and pricing page.
Run a small-volume test with representative production requests and record output format, latency, token usage, and errors.
Increase concurrency in stages while monitoring 429s, 5xx errors, average latency, and actual charges in call logs.
Keep a tested fallback model ID when migrating instead of switching all production traffic at once.

Who was this launch for?

Teams selecting a main Flash model for coding, complex multimodal work, or multi-step agents.
Applications that need high-throughput classification, structured extraction, document parsing, or subagents.
Projects using Gemini 3.5 Flash, Gemini 3.1 Flash-Lite, or an earlier Flash model and evaluating a migration.

A stable existing workload does not need an immediate migration. Compare quality, latency, tool behavior, and total cost with your own prompts and data before replacing a production model.

Frequently asked questions

Are Gemini 3.6 Flash and Gemini 3.5 Flash-Lite production models?

Yes. Google lists gemini-3.6-flash and gemini-3.5-flash-lite as GA and production-ready. Availability for a specific LaoZhang API token group must still be confirmed in the console.

Which model should I choose?

Between the two models here, test Gemini 3.6 Flash first for complex agents, coding, and multimodal tasks, and Gemini 3.5 Flash-Lite first for high-volume extraction, classification, document parsing, and cost-sensitive workloads. For a new project, also test Gemini 3.8 Flash, which launched after this announcement. Decide on measured quality, latency, and total cost, not on the model name alone.

What is the API endpoint for these models?

For the LaoZhang API OpenAI-compatible interface, use the base URL https://api.laozhang.ai/v1; text chat usually calls POST /v1/chat/completions. See the OpenAI SDK guide for full configuration.

Can I migrate existing Gemini code by changing only the model ID?

Do not switch all traffic without testing. Change the model ID in a small-volume test and check tool calls, structured outputs, thinking behavior, token usage, and latency. If you use Google’s native API, also review Google’s latest-model migration guide; native parameter rules should not be assumed to match LaoZhang API’s OpenAI-compatible protocol.

How do I enable or disable thinking?

Gemini 3 models control thinking through request parameters. Some gateway aliases are not official Google model IDs. Use only the parameters or aliases currently displayed by the LaoZhang API console instead of appending a suffix yourself.

Where can I find current pricing?

Google’s upstream pricing and LaoZhang API gateway pricing are separate billing contracts. Current LaoZhang API prices are in the model and pricing catalog; your token groups and actual charges are on the console model and pricing page and in call logs.

​What is the difference between the two models?

​Correct model IDs

​How do I call the models through LaoZhang API?

​Who was this launch for?

​Frequently asked questions

​Are Gemini 3.6 Flash and Gemini 3.5 Flash-Lite production models?

​Which model should I choose?

​What is the API endpoint for these models?

​Can I migrate existing Gemini code by changing only the model ID?

​How do I enable or disable thinking?

​Where can I find current pricing?

​Sources and related links