queries: 200  mean query tokens: 10.1
system prompt tokens: 38
chunk= 256 k= 3 chunks=2057 context=  713.6 prompt_in=  767.7
chunk= 256 k= 5 chunks=2057 context= 1213.1 prompt_in= 1267.2
chunk= 256 k=10 chunks=2057 context= 2391.4 prompt_in= 2445.4
chunk= 512 k= 3 chunks=960 context= 1328.1 prompt_in= 1382.2
chunk= 512 k= 5 chunks=960 context= 2285.3 prompt_in= 2339.4
chunk= 512 k=10 chunks=960 context= 4728.9 prompt_in= 4782.9
chunk=1024 k= 3 chunks=504 context= 2567.0 prompt_in= 2621.1
chunk=1024 k= 5 chunks=504 context= 4143.1 prompt_in= 4197.2
chunk=1024 k=10 chunks=504 context= 8511.9 prompt_in= 8566.0

Cost per 1,000 queries (USD), output=250 tokens assumed; query embedding (10 tokens, approx) included
chunk  k  prompt_in   gpt-5-nano   gpt-5-mini      gpt-5.4  ctx_share_of_gpt-5-mini
  256  3       768       0.1386       0.6921       5.6694  25.8%
  256  5      1267       0.1636       0.8170       6.9181  37.1%
  256 10      2445       0.2225       1.1116       9.8638  53.8%
  512  3      1382       0.1693       0.8457       7.2057  39.3%
  512  5      2339       0.2172       1.0851       9.5988  52.7%
  512 10      4783       0.3393       1.6959      15.7075  69.7%
 1024  3      2621       0.2313       1.1555      10.3029  55.5%
 1024  5      4197       0.3101       1.5495      14.2432  66.8%
 1024 10      8566       0.5285       2.6417      25.1653  80.6%

Indexing: 512-token chunks=960 total tokens=454032 embed cost text-embedding-3-small=$0.0091