Unverified50% confidenceFactExact time
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.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/16/2026
First Seen
Valid until: 1/14/2027
Sources
Related Entities
Related Claims
UnverifiedThe thinking_level parameter of Gemini 4 Argon has three levels: low, medium, high, with medium as the default.77% similarUnverifiedGemini 4 Argon scored 87% on the Embodied Reasoning (ERQA) benchmark, relying entirely on pure reasoning ability without using task-specific fine-tuning or external tool assistance.66% similarUnverifiedGemini 4 Argon emphasizes achieving 52.8% at default effort, and the implied message is that users can get near-top-tier quality without manually raising the reasoning level.63% similarUnverifiedIn the legal work test, Argon scored 19.6% and Fable scored 6.7%, giving Argon nearly a threefold advantage.61% similarUnverifiedThe Bilibili uploader, through the Devin Max plan, invoked Argon at its highest tier and used the same prompt to build six large applications, compiling, testing and reviewing the code one by one; the six applications had a total score of 60, and Argon got 32.6 points (about 54%).61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1044263API
curl https://kongchang.com/api/v1/knowledge/claims/1044263MCP
get_claim(id=1044263)