---
title: "Is Gemini 2.5 Pro Still Available? Access Limits and Migration (2026)"
description: "Gemini 2.5 isn't deprecated, but since Sept 18 only past users can call it. Triage errors, follow a 4-step migration, and check the Oct 20 Cloud retirement date."
url: https://cdnping.org/en/blog/gemini-2-5-access-restricted
language: en
published: 2026-10-18
category: "Troubleshooting"
tags: ["Gemini","Gemini API","Model migration","Troubleshooting","Model deprecation"]
source: cdnping.org
license: Free to quote and summarize with a link to the url above.
---
# Is Gemini 2.5 Pro Still Available? Access Limits and Migration (2026)

Is Gemini 2.5 Pro still available? Yes, with a condition. The Gemini API changelog states, in its September 18, 2026 entry, that the 2.5 series has not been deprecated and will keep being served, but access is now limited to users who have actually used those models before. If you create a new project or a new key, calls to 2.5 may fail. Google recommends 3.5 Flash-Lite or 3.8 Flash for new projects.

Everything below is current as of October 17, 2026 and comes from official documentation. Defer to the docs for details.

## What did Google actually say?

The September 18, 2026 entry in the [Gemini API changelog](https://ai.google.dev/gemini-api/docs/changelog) says, in essence: access to the 2.5 models is restricted to users who have used them in the past; the 2.5 models are not deprecated and will continue to be available through the API until further notice; and new projects should use 3.5 Flash-Lite or 3.8 Flash.

The [Gemini API deprecations page](https://ai.google.dev/gemini-api/docs/deprecations) lists deprecation and shutdown schedules per model. Check it for the current status of the 2.5 series. The changelog's wording is that 2.5 is not deprecated and remains available through the API until further notice.

However, Google Cloud's [model versions documentation](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions) lists the same retirement date for all three models in its "available for at least 12 months after release" table: October 20, 2026 (2.5 Pro and 2.5 Flash were released on June 17, 2025; 2.5 Flash-Lite on July 22, 2025). The two documents disagree: one says "not deprecated, still served," the other gives a retirement date. If you run on Google Cloud's platform, plan around the Cloud documentation and verify actual behavior around October 20. If you call the Gemini API directly, treat the changelog as authoritative, but keep a migration plan ready.

## What errors will I see?

Google has not published a specific error code or message for this restriction. The table below classifies symptoms by what you observe. It is not official guidance.

| Symptom | Likely cause | What to do first |
|---|---|---|
| Model not found (404-type) | Model name is misspelled, or the project/key has no access to 2.5 | Check the model name; test with a newer recommended model |
| Permission denied (403-type) | New project or new key is outside the "has used 2.5" group | Switch models; follow the official docs and console messages |
| Timeouts, connection resets, intermittent failures | Network problem or an outage on the API itself | Check API status, then run a network self-check |

To tell whether the API as a whole is down, start with [Gemini API status](/en/status/gemini). We probe the official API endpoint from Cloudflare's edge every 5 minutes. If the endpoint is healthy and only 2.5 fails while a newer model works, you are almost certainly looking at a model access issue rather than a network issue. For network-layer troubleshooting, see the [developer network self-check checklist (2026 edition)](/en/blog/developer-network-self-check-2026). For how to read status pages, [How to read the OpenAI and Claude status pages](/en/blog/how-to-read-ai-status-pages) covers the same approach.

## How do I migrate? Four steps

**1. Find every place that hard-codes 2.5.** Search your code and config for `gemini-2.5`:

```bash
grep -rn "gemini-2.5" --include="*.py" --include="*.ts" --include="*.js" \
  --include="*.json" --include="*.yaml" --include="*.yml" --include=".env*" .
```

Old tutorials, default configs in open-source projects, and environment variables in CI are the easiest to miss.

**2. Pick a replacement model.** See the table below. In the changelog, Google only gives one recommendation: "new projects should use 3.5 Flash-Lite or 3.8 Flash." The mapping in this table is our suggestion by use case, not an official one-to-one mapping.

| You used | Try first | Why |
|---|---|---|
| gemini-2.5-flash-lite (classification, formatting, lightweight tasks) | 3.5 Flash-Lite | The officially recommended lightweight tier |
| gemini-2.5-flash (everyday general use) | 3.8 Flash | The officially recommended Flash tier |
| gemini-2.5-pro (complex reasoning) | Benchmark against 3.8 Flash first | Google names no direct replacement for 2.5 Pro; decide based on your quality bar |

For the exact model ID spelling, use the Gemini API model list. Don't type it from memory.

**3. Change the configuration.** Common places: the `model` parameter in SDK calls, the model variable in `.env`, and the "Model" or "Model name" field in client tool settings. Menu locations differ by tool, so check each tool's documentation.

**4. Regression-test on your own samples before going live.** Output style, length, latency, and price can all change with the model. Run 20 to 50 real requests through both models and compare them, focusing on format adherence and response time. For how to break latency down, see [AI API latency breakdown](/en/blog/api-latency-connect-rtt-ttft).

## Should I also check pricing and other deprecations?

Yes. Gemini had other changes around the same time. The October 8 changelog says `gemini-3.7-flash` is deprecated and forwarded to 3.8 Flash, and `gemini-3.5-flash` is deprecated and forwarded to 3.6 Flash. The impact is covered in [Gemini 3.7 Flash and 3.5 Flash deprecated, requests auto-forwarded](/en/blog/gemini-flash-deprecation-auto-routing). For pricing, see the [Gemini API price increase timeline](/en/blog/gemini-api-price-increase-dec-2026). Read both before moving to 3.8 Flash so you don't migrate and then get hit by a price increase.

## How can I catch changes like this earlier?

One line in an official changelog can change what you're allowed to call. Our [changes tracker](/en/changes) monitors vendors' official changelogs and pricing pages, records line-by-line additions and removals, and offers an RSS feed. The [model watch](/en/models) records models added, removed, or repriced as seen through OpenRouter's public API. For a fuller approach, see [How to find out right away when LLM API prices change](/en/blog/track-llm-price-and-model-changes).

## Summary and checklist

- 2.5 is not deprecated, but since September 18 access is limited to users who have used it.
- The Cloud documentation gives an October 20 retirement date that conflicts with the Gemini API docs. Verify against the platform you actually use.
- For new projects, use 3.5 Flash-Lite or 3.8 Flash directly.
- Search for `gemini-2.5`, change the model fields, and regression-test before switching.

## Sources

- [Gemini API changelog](https://ai.google.dev/gemini-api/docs/changelog) (entries of 2026-09-18 and 2026-10-08)
- [Gemini API deprecations page](https://ai.google.dev/gemini-api/docs/deprecations)
- [Google Cloud model versions documentation](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions)