Gemini 2.5 Flash and 2.5 Pro withdrawn

The three Gemini 2.5 Flash and 2.5 Pro model IDs were withdrawn on November 12, 2026, and requests for them now return 503. Google has deprecated both models. Switch to the Gemini 3.x model in the table below; the endpoint, API key, and request format stay the same.

Published: November 12, 2026
Last checked: November 12, 2026
Status: Withdrawn

Key facts

Item	Details
Withdrawn models	`gemini-2.5-flash`, `gemini-2.5-pro`, `gemini-2.5-flash-nothinking`
Withdrawn on	November 12, 2026
What you see	HTTP 503 with `no available channels for model gemini-2.5-flash`
Not affected	`gemini-2.5-flash-lite`, `gemini-2.5-flash-image` (Nano Banana), Gemini 3.x and Gemini 4
Endpoints	Unchanged: OpenAI-compatible `/v1/chat/completions` and Gemini-native `/v1beta/models/{model}:generateContent`

Which model to use instead

Old model	New model	When to choose it
`gemini-2.5-flash`	`gemini-3.5-flash-lite`	About the same price as `gemini-2.5-flash`; chat, classification, extraction, and batch jobs
`gemini-2.5-flash`	`gemini-3.8-flash`	Better results on code, multimodal input, and multi-step tasks
`gemini-2.5-pro`	`gemini-3.8-flash`	Lower price than `gemini-2.5-pro`; fits most reasoning and long-document work
`gemini-2.5-pro`	`gemini-3.1-pro-preview`	Complex reasoning and long-document analysis
`gemini-2.5-flash-nothinking`	`gemini-3.5-flash-lite`	Low-latency calls; default thinking level `minimal`

Google also names gemini-3.5-flash-lite and gemini-3.8-flash as the official replacements for these two models. Prices and token groups are in the model catalog; price comparisons on this page use prices as of November 12, 2026.

Who is affected

Scope	Details
Affected	Requests with `model` set to any of the three IDs above, through the OpenAI-compatible or the Gemini-native endpoint
Not affected	`gemini-2.5-flash-lite` stays available; the image model `gemini-2.5-flash-image`; Gemini 3.x and Gemini 4 text models

In the days before the withdrawal, many requests to these models returned HTTP 500 (with code 403) or 429. Failed requests were not charged. Requests now fail at once with 503 and are not routed to another model.

How to migrate

Change the model ID

Set model to the Gemini 3.x model from the table. The endpoint and API key stay the same.

Test with your own prompts

Check answer quality, thinking behavior, tool calls, structured output, and token usage.

Confirm price and billing

Check the price in the model catalog, make sure your token group can call the new model on the token page, and after switching confirm the model name and charge in your call logs.

Example with the OpenAI-compatible endpoint:

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-3.5-flash-lite",
    "messages": [
      {"role": "user", "content": "Turn these meeting notes into three action items"}
    ]
  }'

For the Gemini-native endpoint, change the model name in the path, for example /v1beta/models/gemini-3.5-flash-lite:generateContent. See Gemini native API for the full request.

FAQ

What happens if I still call gemini-2.5-flash?

The request returns HTTP 503 with a message that the model has no available channel, for example Current group default has no available channels for model gemini-2.5-flash. It is not charged and not routed to another model, so change model yourself.

Google retires them on December 1. Why are they gone already?

Google’s deprecations page lists December 1, 2026 as the Gemini API shutdown date for gemini-2.5-flash and gemini-2.5-pro; on Vertex AI they retired on October 20, 2026. From November 9, available capacity for both models kept shrinking and most requests failed, so LaoZhang API withdrew them early instead of letting requests wait more than ten seconds before failing.

Can I still use gemini-2.5-flash-lite?

Yes. Google’s deprecations page lists no shutdown date for gemini-2.5-flash-lite, and LaoZhang API keeps serving it at the same price.

Will my cost change if I switch to gemini-3.5-flash-lite?

Barely. As of November 12, 2026, gemini-3.5-flash-lite has the same input price as gemini-2.5-flash and almost the same output price, and gemini-3.8-flash costs less than gemini-2.5-pro. Current prices are in the model catalog.

How do I control thinking on Gemini 3.x?

Gemini 3 models take the thinking level as a request parameter, and the default differs by model; Gemini 3.5 Flash-Lite defaults to minimal. Do not add suffixes such as -thinking or -nothinking; use the model IDs listed in the model catalog.

Will gemini-2.5-flash or gemini-2.5-pro come back?

There is no plan to bring them back. Google has deprecated both models; if that changes, this page and the announcement index will be updated.

Google: Gemini API deprecations and shutdown dates, Vertex AI model versions and retirement dates
LaoZhang API: the withdrawn status is based on live calls on November 12, 2026; current models, prices, and charges are in the model catalog and your call logs
Related docs: Gemini native API, Gemini 2.5 text models unstable (October notice), How to check whether a model is available

​Key facts

​Which model to use instead

​Who is affected

​How to migrate

​FAQ

​What happens if I still call gemini-2.5-flash?

​Google retires them on December 1. Why are they gone already?

​Can I still use gemini-2.5-flash-lite?

​Will my cost change if I switch to gemini-3.5-flash-lite?

​How do I control thinking on Gemini 3.x?

​Will gemini-2.5-flash or gemini-2.5-pro come back?

​Sources and related docs