Back to rankings

Gemini API: Gemini 2.5 Pro and 2.5 Flash Deprecated, Auto-Routing Starts December 1

Gemini CLIView original changelog

Google deprecated Gemini 2.5 Pro and Gemini 2.5 Flash, which are no longer selectable in Google AI Studio. From December 1, 2026, API requests to gemini-2.5-pro automatically route to Gemini 3.8 Flash, and requests to gemini-2.5-flash route to Gemini 3.5 Flash-Lite, each billed at the replacement model's rates. Developers should update their model strings before then and review the breaking changes around function calling, thinking controls and sampling parameters.

Key Takeaways

  • Gemini 2.5 Pro and 2.5 Flash are deprecated and no longer selectable in Google AI Studio as of November 2, 2026.
  • Automatic routing begins December 1, 2026, so requests keep working but are served by different models.
  • gemini-2.5-pro routes to gemini-3.8-flash, which moves Pro users to a Flash-tier model billed at Gemini 3.8 Flash rates.
  • gemini-2.5-flash routes to gemini-3.5-flash-lite, billed at Gemini 3.5 Flash-Lite rates.
  • thinking_level replaces thinking_budget, so code that sets a numeric thinking budget must be updated.
  • Function calling gets stricter with thought_signature validation, and removed sampling parameters and prefilled model turns can break existing prompts, so testing before the cutover is essential.

What Changed

Google deprecated two of its long-running Gemini API models on November 2, 2026. Both gemini-2.5-pro and gemini-2.5-flash are no longer selectable in Google AI Studio, and both now have a fixed cutover date. On December 1, 2026, Google will begin routing API requests for these models automatically to newer replacements instead of serving the 2.5 models.

Where Requests Will Be Routed

The two models route to different replacements, so teams running both need to plan each migration separately.

Gemini 2.5 Pro

Requests to gemini-2.5-pro will be routed to gemini-3.8-flash starting December 1, 2026. Billing follows Gemini 3.8 Flash rates from that date. Developers who want to avoid the automatic switch should change their model string to gemini-3.8-flash ahead of the cutover. Google points to the Gemini 3.8 migration guide for the details.

Gemini 2.5 Flash

Requests to gemini-2.5-flash will be routed to gemini-3.5-flash-lite, also starting December 1, 2026, and billed at Gemini 3.5 Flash-Lite rates. As with Pro, the recommended action is to update the model string before the cutover and follow the Gemini 3.5 Flash-Lite migration guide.

Breaking Changes to Plan For

Moving off the 2.5 generation is not only a model-name swap. Google lists several breaking changes that apply to both migrations:

  • Stricter function calling behavior, including thought_signature validation.
  • thinking_level replaces thinking_budget for controlling how much the model reasons.
  • Certain sampling parameters are removed.
  • Prefilled model turns are no longer supported.

Applications that pin sampling settings, rely on a numeric thinking budget, or seed conversations with a prefilled model reply will need code changes, not just a new model string.

Why It Matters for Gemini Developers

Because the routing is automatic, teams that do nothing will still be moved on December 1, but with new pricing and with the behavior differences above. Pro users in particular move to a Flash-tier model, so cost and quality characteristics will change. Testing the new models against real prompts and tool-calling flows before the cutover is the safest path. Developers using Gemini through other tools can check which model those tools target, since the underlying API models are changing.