Acropolis

GPT-6 Luna

OpenAIGrade A. Evidence grade A of A to D: strongest.ProprietaryReleased 22 Sept 2026Within-domain variance
Compare with peers

Domain profile

Median within-board percentile by capability domain across boards with sufficient peer overlap.

  • Agentic (12 boards): 26 (scale 0 to 100)Agentic (12 boards)26
  • Coding (8 boards): 39 (scale 0 to 100)Coding (8 boards)39
  • Finance (4 boards): 18 (scale 0 to 100)Finance (4 boards)18
  • General text (3 boards): 61 (scale 0 to 100)General text (3 boards)61
  • Legal (3 boards): 16 (scale 0 to 100)Legal (3 boards)16
  • Science (2 boards): 70 (scale 0 to 100)Science (2 boards)70
  • Search (2 boards): 50 (scale 0 to 100)Search (2 boards)50
  • Vision (2 boards): 56 (scale 0 to 100)Vision (2 boards)56

Benchmark appearances

Best verified run per board, with confidence intervals and Pareto frontier status.

BenchmarkRankScoreCostStatusObserved
LMArena Search (style control)

LMArena · Overall · max

#12 / 42
1,211[1,204 to 1,218]UnknownStandard9 Nov 2026
EMB

Vals AI benchmarks · max

#19 / 38
70%[65% to 74.9%]$0.14Standard15 Nov 2026
Vibe Code Bench

Vals AI benchmarks · max · OpenHands

#21 / 38
81.6%[75% to 88.3%]$1.35Standard15 Nov 2026
FrontierCode

Epoch AI Benchmarking Hub · max · Codex CLI

#22 / 51
42.4%UnknownStandardNot stated
LMArena WebDev

LMArena · Image to webdev · max

#26 / 64
1,567[1,559 to 1,575]UnknownStandard15 Nov 2026
Code Migration

Vals AI benchmarks · max

#27 / 38
45.5%[37.2% to 53.7%]$0.71Standard15 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · max

#30 / 38
13.6%[10.7% to 16.6%]$0.35Standard15 Nov 2026
LMArena Agent

LMArena · Overall · max

#31 / 64
0[0 to 0]$0.08Standard15 Nov 2026
LMArena Search

LMArena · Overall · max

#31 / 42
1,160[1,153 to 1,167]UnknownStandard9 Nov 2026
Vals Index

Vals AI benchmarks · max

#31 / 38
52.3%[50.1% to 54.5%]$0.45Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · max

#32 / 38
56.8%[52.7% to 61%]$0.12Standard15 Nov 2026
Legal Research

Vals AI benchmarks · max

#32 / 38
30.3%[24% to 36.5%]$0.44Standard15 Nov 2026
Tax Agent

Vals AI benchmarks · max

#32 / 38
19.9%[14.3% to 25.6%]$0.19Standard15 Nov 2026
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · max

#33 / 38
4.2%[1% to 7.4%]$0.30Standard15 Nov 2026
LMArena WebDev

LMArena · Webdev HTML · max

#37 / 155
1,595[1,584 to 1,605]UnknownStandard15 Nov 2026
LMArena Agent

LMArena · Overall · medium

#41 / 64
-0[-0 to -0]$0.02Standard15 Nov 2026
LMArena Vision (style control)

LMArena · Chinese · max

#46 / 135
1,304[1,269 to 1,339]UnknownStandard9 Nov 2026
LMArena WebDev

LMArena · Overall · max

#46 / 156
1,577[1,571 to 1,583]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev · max

#46 / 156
1,577[1,571 to 1,583]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev react · max

#48 / 140
1,574[1,567 to 1,580]UnknownStandard15 Nov 2026
LMArena Agent

LMArena · Overall · low

#50 / 64
-0.1[-0.1 to -0.1]$0.01On frontier. Evidence grade A of A to D: strongest.15 Nov 2026
GPQA diamond

Epoch AI Benchmarking Hub · max

#53 / 338
90.5%[86.8% to 94.1%]UnknownStandard22 Sept 2026
LMArena Vision (style control)

LMArena · Homework · max

#62 / 118
1,271[1,251 to 1,291]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · Diagram · max

#66 / 123
1,266[1,249 to 1,282]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · OCR · max

#67 / 123
1,248[1,238 to 1,258]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · Overall · max

#68 / 167
1,235[1,226 to 1,244]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Expert · max

#69 / 371
1,486[1,479 to 1,494]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Creative writing vision · max

#69 / 106
1,217[1,190 to 1,245]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · Chinese · max

#70 / 135
1,274[1,239 to 1,309]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Industry mathematical · max

#72 / 402
1,463[1,454 to 1,472]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Japanese · max

#75 / 288
1,420[1,403 to 1,437]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Polish · max

#78 / 243
1,449[1,433 to 1,465]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Homework · max

#78 / 118
1,252[1,232 to 1,271]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · English · max

#78 / 167
1,227[1,212 to 1,241]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Coding · max

#80 / 416
1,501[1,495 to 1,507]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry software and IT services · max

#80 / 421
1,488[1,483 to 1,494]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Humor · max

#80 / 102
1,194[1,156 to 1,233]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Math · max

#82 / 404
1,449[1,440 to 1,458]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Spanish · max

#82 / 304
1,447[1,435 to 1,459]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Instruction following · max

#86 / 421
1,436[1,431 to 1,442]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Korean · max

#86 / 289
1,400[1,383 to 1,417]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Diagram · max

#86 / 123
1,235[1,219 to 1,251]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · OCR · max

#87 / 123
1,221[1,211 to 1,231]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · Overall · max

#88 / 167
1,211[1,202 to 1,220]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · Creative writing vision · max

#89 / 106
1,190[1,163 to 1,217]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · French · max

#91 / 301
1,466[1,454 to 1,478]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts · max

#91 / 421
1,464[1,459 to 1,469]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Humor · max

#92 / 102
1,168[1,129 to 1,206]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Industry entertainment and sports and media · max

#93 / 419
1,410[1,404 to 1,416]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry writing and literature and language · max

#93 / 420
1,423[1,418 to 1,429]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Exclude ties · max

#94 / 421
1,441[1,436 to 1,447]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · German · max

#94 / 320
1,441[1,424 to 1,457]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Non English · max

#94 / 421
1,430[1,425 to 1,434]UnknownStandard11 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · max · Codex CLI

#95 / 145
18.3$0.52On frontier. Evidence grade A of A to D: strongest.23 Sept 2026
LMArena Text (style control)

LMArena · Longer query · max

#95 / 399
1,449[1,443 to 1,454]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · English · max

#95 / 167
1,204[1,189 to 1,218]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts English · max

#97 / 419
1,467[1,461 to 1,472]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Creative writing · max

#98 / 419
1,408[1,402 to 1,414]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Overall · max

#99 / 421
1,441[1,437 to 1,445]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Chinese · max

#100 / 395
1,477[1,469 to 1,485]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Russian · max

#102 / 385
1,437[1,430 to 1,444]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · English · max

#104 / 421
1,447[1,442 to 1,452]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry legal and government · max

#107 / 393
1,445[1,437 to 1,453]UnknownStandard11 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · max

#108 / 299
59.3%UnknownStandardNot stated
LMArena Text

LMArena · Expert · max

#108 / 371
1,437[1,430 to 1,445]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry life and physical and social science · max

#108 / 419
1,453[1,447 to 1,460]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Multi turn · max

#108 / 419
1,441[1,435 to 1,448]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry business and management and financial operations · max

#109 / 414
1,439[1,433 to 1,446]UnknownStandard11 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · xhigh · Codex CLI

#115 / 145
14$0.39On frontier. Evidence grade A of A to D: strongest.23 Sept 2026
LMArena Text

LMArena · Japanese · max

#116 / 288
1,366[1,349 to 1,383]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Instruction following · max

#125 / 421
1,403[1,398 to 1,409]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry mathematical · max

#128 / 402
1,425[1,416 to 1,434]UnknownStandard11 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · high · Codex CLI

#129 / 145
9$0.13On frontier. Evidence grade A of A to D: strongest.23 Sept 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · xhigh

#137 / 299
41.9%UnknownStandardNot stated
Bug Hunt Bench

Bug Hunt Bench · medium · Codex CLI

#138 / 145
4$0.06On frontier. Evidence grade A of A to D: strongest.23 Sept 2026
Bug Hunt Bench

Bug Hunt Bench · low · Codex CLI

#138 / 145
4$0.20Standard23 Sept 2026
LMArena Text

LMArena · Coding · max

#141 / 416
1,431[1,425 to 1,437]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Korean · max

#144 / 289
1,347[1,330 to 1,364]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Math · max

#144 / 404
1,411[1,402 to 1,420]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Polish · max

#144 / 243
1,382[1,366 to 1,398]UnknownStandard11 Nov 2026
LMArena Text

LMArena · German · max

#145 / 320
1,384[1,367 to 1,400]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Longer query · max

#146 / 399
1,403[1,397 to 1,408]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry writing and literature and language · max

#147 / 420
1,382[1,376 to 1,387]UnknownStandard11 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · high

#149 / 299
31.4%UnknownStandardNot stated
LMArena Text

LMArena · French · max

#150 / 301
1,405[1,393 to 1,417]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry software and IT services · max

#152 / 421
1,422[1,417 to 1,427]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Spanish · max

#155 / 304
1,395[1,384 to 1,407]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry medicine and healthcare · max

#155 / 391
1,428[1,420 to 1,437]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Russian · max

#156 / 385
1,388[1,381 to 1,395]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Creative writing · max

#157 / 419
1,359[1,353 to 1,365]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Hard prompts · max

#158 / 421
1,407[1,403 to 1,412]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry entertainment and sports and media · max

#161 / 419
1,359[1,353 to 1,364]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Chinese · max

#165 / 395
1,422[1,414 to 1,430]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Multi turn · max

#165 / 419
1,393[1,386 to 1,399]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Hard prompts English · max

#166 / 419
1,407[1,401 to 1,412]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Non English · max

#167 / 421
1,380[1,375 to 1,385]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry legal and government · max

#168 / 393
1,392[1,384 to 1,400]UnknownStandard11 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · medium

#169 / 299
18.1%UnknownStandardNot stated
LMArena Text

LMArena · Industry business and management and financial operations · max

#171 / 414
1,382[1,376 to 1,388]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Exclude ties · max

#172 / 421
1,366[1,360 to 1,371]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Overall · max

#173 / 421
1,388[1,384 to 1,393]UnknownStandard11 Nov 2026
LMArena Text

LMArena · English · max

#178 / 421
1,391[1,386 to 1,396]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry life and physical and social science · max

#179 / 419
1,387[1,380 to 1,393]UnknownStandard11 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · low

#210 / 299
4.6%UnknownStandardNot stated
LMArena Text

LMArena · Industry medicine and healthcare · max

#213 / 391
1,349[1,341 to 1,358]UnknownStandard11 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · none

#281 / 299
0%UnknownStandardNot stated

Reasoning effort scaling

Same benchmark and harness measured across multiple reasoning effort settings.

  • ARC-AGI-2

    Epoch AI Benchmarking Hub

    Monotone. Evidence grade A of A to D: strongest.
    none

    0.0

    low

    4.6

    medium

    18.1

    high

    31.4

    xhigh

    41.9

    max

    59.3

  • Bug Hunt Bench

    Bug Hunt Bench · Codex CLI

    Monotone. Evidence grade A of A to D: strongest.
    low

    4.0

    $0.20

    medium

    4.0

    $0.06

    high

    9.0

    $0.13

    xhigh

    14.0

    $0.39

    max

    18.3

    $0.52

  • LMArena Agent

    LMArena

    Monotone. Evidence grade A of A to D: strongest.
    low

    -0.1

    $0.01

    medium

    -0.0

    $0.02

    max

    0.0

    $0.08

Closest head-to-head rivals

Direct pairwise win rates for GPT-6 Luna across shared benchmark boards.