← Back to GeminiPulse Home

Gemini 2.5 Migration Series • Part 1 of 3

Gemini 2.5 Sunset & Auto-Routing: What Changed on November 2, 2026 & The Dec 1 Cutover

Published November 2, 2026 • Primary Google Sources: Google AI for Developers Changelog (Nov 2, 2026) & Gemini API Deprecations Documentation
⚠️ Critical Developer Action Item:

On November 2, 2026, Google officially deprecated gemini-2.5-pro and gemini-2.5-flash. Both models have been removed from the Google AI Studio model picker. Starting December 1, 2026, all active API requests pointing to legacy 2.5 model strings will automatically route to modern 3.x models. While this brings immediate token cost savings of up to 80%, unmanaged migrations risk failing due to breaking API requirements around tool validation and thinking parameters.

1. The Model Deprecation Matrix & Auto-Routing Paths

Deprecated Model Status as of Nov 2, 2026 Automatic Routing Target (Dec 1, 2026) Pricing Impact (per 1M tokens)
gemini-2.5-pro Removed from AI Studio; API deprecated gemini-3.8-flash Drops from $3.50 / $10.50 down to $0.75 / $3.75 (-78% / -64%)
gemini-2.5-flash Removed from AI Studio; API deprecated gemini-3.5-flash-lite Drops from $0.35 / $1.05 down to $0.10 / $0.40 (-71% / -62%)

2. Why Google Is Enforcing Auto-Routing

Unlike hard-shutdown deprecations that instantly return 404 Not Found or 410 Gone, Google has engineered an automatic backend fallback router for the December 1 cutover. Any legacy client that fails to update its model string will continue resolving requests, but with two monumental shifts:

3. The December 1 Cutover Timeline

The Takeaway:

Auto-routing prevents catastrophic service outages for unmaintained endpoints, but developers should not rely on it as a permanent strategy. Gemini 3.x enforces stricter schema validation and a different thinking control schema that must be addressed explicitly in client code.