thinking_level
A budget tier governing the internal reasoning a large model is allowed to perform before giving its final answer. The higher the level, the more tokens the model generates during the implicit thinking stage. Reasoning quality is usually better, but both first-token latency and total output tokens increase.
Timeline (last 90 days)
Argon's thinking_level parameter has three tiers: low, medium (default), and high; reasoning tokens are billed at the output price, even if the reasoning process is not shown to the user.
Argon's low tier reported zero reasoning tokens across 39 calls, with all outputs only 5-11 tokens; the medium tier had a median output of 383 tokens; the high tier had a median output of 1,455 tokens.
The thinking_level parameter of Gemini 4 Argon has three levels: low, medium, high, with medium as the default.
All Facts (3)
Argon's thinking_level parameter has three tiers: low, medium (default), and high; reasoning tokens are billed at the output price, even if the reasoning process is not shown to the user.
50%UnverifiedArgon's low tier reported zero reasoning tokens across 39 calls, with all outputs only 5-11 tokens; the medium tier had a median output of 383 tokens; the high tier had a median output of 1,455 tokens.
50%UnverifiedThe thinking_level parameter of Gemini 4 Argon has three levels: low, medium, high, with medium as the default.
50%