In addition to the base model (e.g. gemini-2.5-pro), the project provides special features through suffix variants:
- -search variant(Example: gemini-2.5-pro-search):
Activate Google search enhancements for scenarios that require real-time web information retrieval, such as news summary generation - -nothinking variant::
Reduced model inference steps, sacrificing part of the answer quality for faster response time, suitable for real-time dialog scenarios - -maxthinking variant::
Increased inference budget to 32,768 tokens for complex problems requiring deep analysis, such as academic paper interpretation
Developers can directly specify these variant names during API calls, and the system will automatically adapt the corresponding inference parameters. This design preserves the original capabilities of Gemini while providing optimization options for different scenarios.
This answer comes from the articlegeminicli2api: Proxy tool to convert Gemini CLI to OpenAI-compatible APIsThe
May not be reproduced without permission:AI productivity tools " What special model variants are supported by geminicli2api? What does each one do?
