The image above show the exact SAME prompt, with SAME context, the unique difference is the model. The sonnet was 26x more expensive, and used 3x more tokens, this does not make any sense.
PS: This was a simple request with really small previous context.
Request ID: b56c316d-6cd2-42dd-85c7-85fe226e193e
1 Like
well as i understand it, each model decides on its own how many and what tool calls it does. and also you might use same model twice, first time 5 tool calls, second time 15 tool calls.
each tool call must have the entire context passed as reference (if im not mistaken) but it is cache read context (usually cheap)
1 Like
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.