{ "model": "nvidia/Nemotron-3.5-Parse", "hardware": "One H200, BF16", "pages_per_trial": 2078, "independent_repeats": 3, "statistic": "Median; observed min/max are not confidence intervals", "selection_metric": "Fully completed pages per inference second", "winners": { "single": { "settings": { "mode": "single", "concurrency": 128, "capacity": 128, "draft": 16, "batch_tokens": 32768, "prefix": true, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 17.777170543750234, "processed_rate": 18.011194729357868, "completed_range": [ 17.63583791664956, 17.908738148240406 ], "processed_range": [ 17.876717654047702, 18.16220491558983 ], "complete_pages": [ 2049, 2051, 2050 ], "tokens": [ 2678232, 2674047, 2680942 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, "double": { "settings": { "mode": "double", "concurrency": 128, "capacity": 192, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 13.995984844135963, "processed_rate": 14.263686368864409, "completed_range": [ 13.635665159920789, 14.507522594142927 ], "processed_range": [ 13.896474841743698, 14.756060670890358 ], "complete_pages": [ 2043, 2039, 2039 ], "tokens": [ 3791781, 3839100, 3819316 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 2337 }, { "fullpage": 2078, "detection": 0, "crop": 2353 }, { "fullpage": 2078, "detection": 0, "crop": 2343 } ] }, "triple": { "settings": { "mode": "triple", "concurrency": 256, "capacity": 256, "draft": 16, "batch_tokens": 65536, "prefix": true, "cache": 8, "memory": 0.9, "keepalive": "pooled" }, "completed_rate": 12.760647624071124, "processed_rate": 12.991977345820576, "completed_range": [ 12.665496310560572, 12.948023577542324 ], "processed_range": [ 12.914083088000426, 13.18275011961438 ], "complete_pages": [ 2041, 2041, 2038 ], "tokens": [ 4192165, 4190074, 4207913 ], "requests": [ { "fullpage": 2078, "detection": 2078, "crop": 2625 }, { "fullpage": 2078, "detection": 2078, "crop": 2633 }, { "fullpage": 2078, "detection": 2078, "crop": 2610 } ] } }, "confirmed_profiles": [ { "settings": { "mode": "single", "concurrency": 1, "capacity": 64, "draft": 32, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 2.6001723809921358, "processed_rate": 2.6382608436043253, "completed_range": [ 2.575941521411921, 2.6236829217680655 ], "processed_range": [ 2.6136750397919783, 2.6621157770674024 ], "complete_pages": [ 2048, 2048, 2048 ], "tokens": [ 2690504, 2690504, 2690504 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "single", "concurrency": 2, "capacity": 2, "draft": 32, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 4.70235904598427, "processed_rate": 4.775905228521658, "completed_range": [ 4.682969565196626, 4.824188615533726 ], "processed_range": [ 4.753888987043766, 4.894855440956583 ], "complete_pages": [ 2046, 2048, 2047 ], "tokens": [ 2708815, 2695983, 2698871 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "single", "concurrency": 4, "capacity": 4, "draft": 32, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 7.726758637907933, "processed_rate": 7.83229485345009, "completed_range": [ 7.507347378031041, 8.179612482828606 ], "processed_range": [ 7.617318286888918, 8.287291437990172 ], "complete_pages": [ 2051, 2050, 2048 ], "tokens": [ 2667442, 2683911, 2695038 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "single", "concurrency": 8, "capacity": 8, "draft": 32, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 11.300312866641013, "processed_rate": 11.443494218752448, "completed_range": [ 11.114101752060728, 11.302029176217788 ], "processed_range": [ 11.260411233925984, 11.445232274941796 ], "complete_pages": [ 2051, 2052, 2052 ], "tokens": [ 2670519, 2674332, 2675976 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "single", "concurrency": 16, "capacity": 16, "draft": 32, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 13.72132422096978, "processed_rate": 13.929121510100247, "completed_range": [ 13.629552372710384, 13.868329751046273 ], "processed_range": [ 13.80897602656859, 14.044049328788576 ], "complete_pages": [ 2052, 2051, 2047 ], "tokens": [ 2664515, 2667117, 2692456 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "single", "concurrency": 32, "capacity": 32, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 16.24820263620281, "processed_rate": 16.454076548747288, "completed_range": [ 15.913943267172057, 16.505446511638127 ], "processed_range": [ 16.139177212876298, 16.70643831036728 ], "complete_pages": [ 2053, 2049, 2052 ], "tokens": [ 2665461, 2678495, 2668148 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "single", "concurrency": 64, "capacity": 64, "draft": 8, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 17.03946058790172, "processed_rate": 17.322895842299303, "completed_range": [ 16.779610418138926, 17.116187503834503 ], "processed_range": [ 17.00050241291696, 17.3753969872829 ], "complete_pages": [ 2044, 2047, 2051 ], "tokens": [ 2703005, 2687559, 2677871 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "double", "concurrency": 128, "capacity": 128, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 13.871913662638613, "processed_rate": 14.116472375594045, "completed_range": [ 13.521083161335971, 14.320910857155429 ], "processed_range": [ 13.772946475125563, 14.573385289504888 ], "complete_pages": [ 2042, 2042, 2040 ], "tokens": [ 3792654, 3811233, 3810040 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 2341 }, { "fullpage": 2078, "detection": 0, "crop": 2340 }, { "fullpage": 2078, "detection": 0, "crop": 2343 } ] }, { "settings": { "mode": "double", "concurrency": 128, "capacity": 128, "draft": 16, "batch_tokens": 32768, "prefix": true, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 13.857578244206481, "processed_rate": 14.101884226964284, "completed_range": [ 13.427124064834844, 14.14685037641301 ], "processed_range": [ 13.650471529709787, 14.375136959504271 ], "complete_pages": [ 2042, 2044, 2045 ], "tokens": [ 3817020, 3819064, 3799365 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 2338 }, { "fullpage": 2078, "detection": 0, "crop": 2342 }, { "fullpage": 2078, "detection": 0, "crop": 2344 } ] }, { "settings": { "mode": "double", "concurrency": 128, "capacity": 192, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 13.995984844135963, "processed_rate": 14.263686368864409, "completed_range": [ 13.635665159920789, 14.507522594142927 ], "processed_range": [ 13.896474841743698, 14.756060670890358 ], "complete_pages": [ 2043, 2039, 2039 ], "tokens": [ 3791781, 3839100, 3819316 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 2337 }, { "fullpage": 2078, "detection": 0, "crop": 2353 }, { "fullpage": 2078, "detection": 0, "crop": 2343 } ] }, { "settings": { "mode": "single", "concurrency": 128, "capacity": 128, "draft": 16, "batch_tokens": 32768, "prefix": true, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 17.777170543750234, "processed_rate": 18.011194729357868, "completed_range": [ 17.63583791664956, 17.908738148240406 ], "processed_range": [ 17.876717654047702, 18.16220491558983 ], "complete_pages": [ 2049, 2051, 2050 ], "tokens": [ 2678232, 2674047, 2680942 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "single", "concurrency": 128, "capacity": 192, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 17.419733802275566, "processed_rate": 17.666279571073023, "completed_range": [ 17.10274000419257, 17.97142015619434 ], "processed_range": [ 17.353268422222733, 18.225774077389868 ], "complete_pages": [ 2049, 2049, 2048 ], "tokens": [ 2681005, 2673131, 2688923 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "triple", "concurrency": 128, "capacity": 192, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 12.187296867472998, "processed_rate": 12.396085604801218, "completed_range": [ 12.009630198553252, 12.6213920745905 ], "processed_range": [ 12.23333899636944, 12.856496436764244 ], "complete_pages": [ 2043, 2040, 2040 ], "tokens": [ 4183215, 4188764, 4191600 ], "requests": [ { "fullpage": 2078, "detection": 2078, "crop": 2621 }, { "fullpage": 2078, "detection": 2078, "crop": 2605 }, { "fullpage": 2078, "detection": 2078, "crop": 2601 } ] }, { "settings": { "mode": "single", "concurrency": 256, "capacity": 256, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 17.6721021229776, "processed_rate": 17.89601764695295, "completed_range": [ 17.107707265642375, 17.85725949121545 ], "processed_range": [ 17.375276489738443, 18.05712176289329 ], "complete_pages": [ 2046, 2052, 2055 ], "tokens": [ 2683495, 2666175, 2660412 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] }, { "settings": { "mode": "triple", "concurrency": 256, "capacity": 512, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": "pooled" }, "completed_rate": 12.648139018891586, "processed_rate": 12.883741608459175, "completed_range": [ 12.063085322992798, 12.983478678422928 ], "processed_range": [ 12.305886745792359, 13.199446523367342 ], "complete_pages": [ 2040, 2037, 2044 ], "tokens": [ 4196553, 4211493, 4171417 ], "requests": [ { "fullpage": 2078, "detection": 2078, "crop": 2604 }, { "fullpage": 2078, "detection": 2078, "crop": 2609 }, { "fullpage": 2078, "detection": 2078, "crop": 2611 } ] }, { "settings": { "mode": "triple", "concurrency": 256, "capacity": 256, "draft": 16, "batch_tokens": 65536, "prefix": true, "cache": 8, "memory": 0.9, "keepalive": "pooled" }, "completed_rate": 12.760647624071124, "processed_rate": 12.991977345820576, "completed_range": [ 12.665496310560572, 12.948023577542324 ], "processed_range": [ 12.914083088000426, 13.18275011961438 ], "complete_pages": [ 2041, 2041, 2038 ], "tokens": [ 4192165, 4190074, 4207913 ], "requests": [ { "fullpage": 2078, "detection": 2078, "crop": 2625 }, { "fullpage": 2078, "detection": 2078, "crop": 2633 }, { "fullpage": 2078, "detection": 2078, "crop": 2610 } ] }, { "settings": { "mode": "single", "concurrency": 512, "capacity": 512, "draft": 16, "batch_tokens": 32768, "prefix": false, "cache": 4, "memory": 0.8, "keepalive": 0 }, "completed_rate": 17.215383924028252, "processed_rate": 17.44201257636797, "completed_range": [ 16.814386036148, 17.33674505992429 ], "processed_range": [ 17.07736763593135, 17.590701286388025 ], "complete_pages": [ 2048, 2046, 2051 ], "tokens": [ 2679913, 2697997, 2675209 ], "requests": [ { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 }, { "fullpage": 2078, "detection": 0, "crop": 0 } ] } ] }