Observation/Feedback #2
Description
Nothing to complain great work @dprevoznik with few LOC. As I am prototyping in this space, browser-use seems to be more cost efficient.
Kernel with AI SDK - gpt-oss-120b ($0.094) - 30~40 secs - input token 350k
Browser-use OSS - gemini-flash-latest ($0.023) - 70~100 secs - input token 90k
For simple use case this is negligible, but thinking about scale is where the concern is.
But I really like that kernel is so quick, If you guys can bring down that token usage there's a real edge. Ping me for more info