ugh seriously losing my mind over this. whenever i route kimi k3 through openrouter or other API gateways, the token usage and tracking just breaks completely.
the reasoning tokens and input/output counts are totally miscalculated, and sometimes it causes my context window calls to just throw weird errors or get cut off mid-way even though i'm well under the 1m limit. like my billing dashboard is showing random spikes and it feels like the router integration isn't parsing the token headers right.
is anyone else dealing with this?? it's driving me crazy trying to run agent tasks. pls look into this asap 😩
ugh seriously losing my mind over this. whenever i route kimi k3 through openrouter or other API gateways, the token usage and tracking just breaks completely.
the reasoning tokens and input/output counts are totally miscalculated, and sometimes it causes my context window calls to just throw weird errors or get cut off mid-way even though i'm well under the 1m limit. like my billing dashboard is showing random spikes and it feels like the router integration isn't parsing the token headers right.
is anyone else dealing with this?? it's driving me crazy trying to run agent tasks. pls look into this asap 😩