Update: After extensive testing, the routing trigger is input token count, not content. Requests with ~12,600 prompt tokens are served by Pro; adding ~73 more tokens causes routing to Flash Lite. The threshold appears to be approximately 12,600–12,700 input tokens. Content, safety settings, API endpoint, and API key are all ruled out as factors.
atanos
2
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| AI Studio API keeps downgrading to Gemini 3.1 Flash Lite | 1 | 153 | November 13, 2026 | |
| Gemini-4-carbon changes modelVersion to Flash Lite at STOP (native HTTPS repro) | 2 | 308 | November 13, 2026 | |
| Gemini PRO unavailable in Google AI Studio for Google One Subscribers | 1 | 245 | October 1, 2026 | |
| Gemini 3.1 pro was good a week ago, but now it's completely down | 12 | 3056 | May 26, 2026 | |
| Gemini 3.8 Flash ignored System Instructions, caused an unexplained API downgrade (3.8 to 2.5), and broke working TS code | 1 | 335 | October 27, 2026 |