Gemini 3.5 Flash is deprecated. Google AI Studio no longer offers it for selection, and requests using gemini-3.5-flash are automatically routed to gemini-3.6-flash. Google recommends updating the model string to 3.6 Flash or 3.8 Flash. A successful request using the old name therefore does not establish that the original model is serving it. Google model documentation.
Deprecation and Endpoint Availability
Google's deprecation schedule says the 3.5 Flash endpoint is not shutting down: it redirects requests to the replacement. This differs from an endpoint that has been switched off.
Make the intended model explicit in your configuration. An old release announcement cannot establish present availability, and a documentation update date does not establish when routing changed.
Replacement Models
| Target | Migration consideration |
|---|---|
gemini-3.6-flash |
Matches the documented automatic-routing destination |
gemini-3.8-flash |
An alternative recommended on the deprecated model page; review its settings before switching |
Gemini 3.6 Flash is listed as stable. Its documented inputs include text, images, video, audio and PDFs, with text output.
Gemini 3.8 Flash is also stable. Its thinking settings are low, medium and high; minimal is unsupported and returns an error. Audit that setting before replacing the model name.
For an existing configuration, the smallest explicit change to the current routing destination is:
- model = "gemini-3.5-flash"
+ model = "gemini-3.6-flash"
This configuration change needs regression checks before deployment. Validate representative prompts, structured responses and tool calls. Record latency, token consumption and failure handling with your own workload; the version number alone cannot establish those results.
SDK Migration
Model selection and SDK migration are separate checks. Google's SDK migration guide replaces Python's google-generativeai with google-genai, and JavaScript's @google/generative-ai with @google/genai. The new libraries use a central client; code using legacy interfaces needs the corresponding changes.
Confirm which provider and API surface your application uses. Third-party providers may apply different aliases and routing rules; verify those with the provider before changing production configuration.
Rankings and Serving Status
A leaderboard entry, a provider model ID and an API routing alias serve different purposes. Compare the exact identifiers and configurations before treating them as the same model. A name appearing on a leaderboard does not establish that its original API endpoint still serves that model.
Use provider documentation to establish service availability and rankings to inform comparison. For further evaluation of the replacement, AI Rank also has a 3.8 Flash pricing, specifications and rankings overview; interpret its dated figures within their stated windows.
Prepared with automated editorial assistance; official documentation checked on October 21, 2026. No independent API call tests were performed.