I’m trying to set the thinking_level configuration to “medium” using the BatchAPI but can’t seem to configure it correctly. Using the client.models.generate_content I get it to work.
This is a screenshot of my code to generate the jsonl file
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Which fields does the Batch API support when using the OpenAI library to do batch inference via Gemini-3? | 0 | 39 | January 15, 2026 | |
| Gemini 2.5 Flash Thinking Tokens using OpenAI API | 16 | 1703 | June 12, 2025 | |
| Context caching - batch api requests | 4 | 274 | November 25, 2025 | |
| Structured output in gemini-2.5-flash-lite batch mode (input file) | 1 | 218 | November 25, 2025 | |
| Thinking_level low causes error in gemini-2.5-flash-lite (steps version of interactions API) | 0 | 17 | May 10, 2026 |