Search-SWE

GPT-6 Luna

reasoning effort max · Last Updated

openai/gpt-6-luna

Task results

Task catalog ↗
Task results for GPT-6 Luna
TaskHardwareAgentProviderTrajectory
agentic-search CPUCodex (v0.164.0)OpenRouter1.02h 0min69,569,763295,47069,124,471$1.41Trace →
agentic-search-vectorless CPUCodex (v0.164.0)OpenRouter0.01h 10min36,321,762229,59436,019,370$0.60Trace →
biomedical-graph-retrieval CPUCodex (v0.164.0)OpenRouter10.044min11,377,906177,29511,146,275$0.23Trace →
budgeted-bimetric-retrieval CPUCodex (v0.164.0)OpenRouter18.236min7,577,611101,5737,437,861$0.14Trace →
code-context-retrieval CPUCodex (v0.164.0)OpenRouter16.71h 35min51,839,738353,68051,348,852$1.23Trace →
evolving-personal-memory CPUCodex (v0.164.0)OpenRouter32.845min15,087,765191,63114,836,036$0.28Trace →
filtered-ann-retrieval CPUCodex (v0.164.0)OpenRouter0.055min9,514,295160,9409,317,119$0.20Trace →
financial-report-qa CPUCodex (v0.164.0)OpenRouter10.654min14,524,405171,72214,286,415$0.26Trace →
icsi-meeting-memory CPUCodex (v0.164.0)OpenRouter0.01h 37min43,594,608306,49543,104,028$1.06Trace →
incremental-vector-search CPUCodex (v0.164.0)OpenRouter0.02h 0min52,081,738211,77451,787,046$0.76Trace →
join-aware-table-retrieval CPUCodex (v0.164.0)OpenRouter40.037min10,151,616150,7729,922,632$0.20Trace →
large-scale-formula-retrieval CPUCodex (v0.164.0)OpenRouter16.72h 0min13,206,754180,31812,987,387$0.25Trace →
long-document-reranking CPUCodex (v0.164.0)OpenRouter33.01h 32min23,608,663197,53823,336,756$0.37Trace →
long-pdf-evidence CPUCodex (v0.164.0)OpenRouter33.31h 22min25,919,795273,19525,580,157$0.59Trace →
long-pdf-evidence-vectorless CPUCodex (v0.164.0)OpenRouter8.31h 9min24,541,032246,67124,138,950$0.59Trace →
memory-constrained-dense-retrieval CPUCodex (v0.164.0)OpenRouter83.040min15,760,687216,78315,493,549$0.30Trace →
narrative-clue-search CPUCodex (v0.164.0)OpenRouter6.22h 0min39,224,828321,14538,358,508$1.05Trace →
negative-constraint-retrieval CPUCodex (v0.164.0)OpenRouter0.040min11,078,228166,73810,864,413$0.22Trace →
paper-metadata-retrieval CPUCodex (v0.164.0)OpenRouter0.01h 4min20,783,784218,73120,507,870$0.37Trace →
reasoning-query-rewriting CPUCodex (v0.164.0)OpenRouter30.02h 0min19,348,336188,14019,050,943$0.32Trace →
rust-bm25-search CPUCodex (v0.164.0)OpenRouter24.133min7,251,291128,5897,018,201$0.16Trace →
scientific-paper-qa CPUCodex (v0.164.0)OpenRouter20.02h 0min58,004,541237,57757,333,274$1.13Trace →
sparse-retrieval-pruning CPUCodex (v0.164.0)OpenRouter0.049min14,483,338177,93814,255,623$0.26Trace →
visual-document-qa CPUCodex (v0.164.0)OpenRouter17.848min17,102,874189,67016,725,088$0.31Trace →
budgeted-candidate-reranking GPUCodex (v0.164.0)OpenRouter7.336min10,478,34785,48710,312,377$0.17Trace →
chunqiu-temporal-retrieval GPUCodex (v0.164.0)OpenRouter1.749min13,553,651184,12413,278,544$0.28Trace →
code-embedding-finetuning GPUCodex (v0.164.0)OpenRouter5.02h 47min19,197,198103,16118,996,060$0.27Trace →
code-reranker-finetuning GPUCodex (v0.164.0)OpenRouter40.93h 7min13,884,69091,08913,725,215$0.20Trace →
composed-image-search GPUCodex (v0.164.0)OpenRouter36.349min6,697,157142,5816,469,145$0.16Trace →
constrained-molecular-search GPUCodex (v0.164.0)OpenRouter15.52h 40min55,719,386200,56755,410,774$0.98Trace →
decision-model-training GPUCodex (v0.164.0)OpenRouter38.33h 31min21,749,702165,55621,531,216$0.33Trace →
document-page-search GPUCodex (v0.164.0)OpenRouter7.01h 3min16,534,852134,62216,319,010$0.26Trace →
instruction-following-embedding GPUCodex (v0.164.0)OpenRouter0.050min11,116,34193,21710,966,393$0.18Trace →
khmer-embedding-training GPUCodex (v0.164.0)OpenRouter14.61h 20min9,512,541119,0969,343,679$0.17Trace →
layout-page-search GPUCodex (v0.164.0)OpenRouter18.81h 4min14,568,292104,32914,393,597$0.22Trace →
legal-stance-retrieval GPUCodex (v0.164.0)OpenRouter9.932min5,445,518102,1855,307,722$0.12Trace →
lightweight-query-retrieval GPUCodex (v0.164.0)OpenRouter0.03h 18min129,909,430487,875129,306,297$2.77Trace →
methodology-inspiration-retrieval GPUCodex (v0.164.0)OpenRouter3.445min13,552,387145,51913,357,183$0.23Trace →
multilingual-vector-postprocessing GPUCodex (v0.164.0)OpenRouter26.41h 4min16,088,069142,56015,888,705$0.26Trace →
protein-text-bidirectional-retrieval GPUCodex (v0.164.0)OpenRouter16.756min9,773,36176,6979,618,692$0.15Trace →
quantity-aware-retrieval GPUCodex (v0.164.0)OpenRouter0.045min12,673,530136,92012,491,690$0.22Trace →
query-encoder-alignment GPUCodex (v0.164.0)OpenRouter28.946min11,669,767143,05211,453,733$0.21Trace →
same-place-image-search GPUCodex (v0.164.0)OpenRouter57.61h 2min9,296,237113,0429,131,007$0.17Trace →
sketch-text-search GPUCodex (v0.164.0)OpenRouter3.51h 24min21,040,279157,33320,783,360$0.32Trace →
spoken-query-search GPUCodex (v0.164.0)OpenRouter13.82h 0min18,913,018190,68118,426,150$0.34Trace →
training-error-tracing GPUCodex (v0.164.0)OpenRouter0.024min7,523,40193,1797,356,046$0.14Trace →
video-moment-search GPUCodex (v0.164.0)OpenRouter22.824min3,774,714108,3853,622,157$0.11Trace →

47 selected evaluations · valid zero scores included

Score: task reward × 100. Tokens: coding usage; cached tokens are included in input.

Cost is estimated from recorded agent tokens at official API rates in USD (2026-11-01; DeepSeek off-peak pricing).