>>108785795For some reason 60k is the sweet spot, even though the actual token utilization looks like this:
usage: {
prompt_tokens: 58020,
completion_tokens: 623,
total_tokens: 58643,
prompt_tokens_details: { cached_tokens: 512 },
completion_tokens_details: { reasoning_tokens: 383 },
prompt_cache_hit_tokens: 512,
prompt_cache_miss_tokens: 57508
},