Acropolis

GPT-6.1 Sol

OpenAIGrade A. Evidence grade A of A to D: strongest.ProprietaryReleased 29 Sept 2026Consistent across benchmarks
Compare with peers

Domain profile

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

  • Agentic (13 boards): 78 (scale 0 to 100)Agentic (13 boards)78
  • Coding (8 boards): 85 (scale 0 to 100)Coding (8 boards)85
  • Finance (5 boards): 60 (scale 0 to 100)Finance (5 boards)60
  • General text (3 boards): 93 (scale 0 to 100)General text (3 boards)93
  • Legal (3 boards): 31 (scale 0 to 100)Legal (3 boards)31
  • Science (2 boards): 98 (scale 0 to 100)Science (2 boards)98
  • Search (2 boards): 90 (scale 0 to 100)Search (2 boards)90
  • Vision (2 boards): 90 (scale 0 to 100)Vision (2 boards)90

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
LMArena Vision (style control)

LMArena · Homework · max

#1 / 118Tied #1
1,357[1,333 to 1,381]UnknownStandard9 Nov 2026
LMArena WebDev

LMArena · Webdev HTML · max

#1 / 155Tied #1
1,796[1,772 to 1,820]UnknownStandard15 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · max

#2 / 299
94.2%UnknownStandardNot stated
LMArena Search (style control)

LMArena · Overall · max

#2 / 42Tied #1
1,248[1,240 to 1,256]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · Homework · max

#3 / 118Tied #1
1,353[1,329 to 1,377]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · Creative writing vision · max

#5 / 106Tied #1
1,303[1,271 to 1,335]UnknownStandard9 Nov 2026
LMArena WebDev

LMArena · Image to webdev · max

#5 / 64
1,694[1,681 to 1,706]UnknownStandard15 Nov 2026
Vending-Bench 2

Epoch AI Benchmarking Hub · Default

#5 / 72
$12,860.61UnknownStandardNot stated
GPQA diamond

Epoch AI Benchmarking Hub · max

#6 / 338Tied #1
95.4%[92.7% to 98.1%]UnknownStandard29 Sept 2026
Code Migration

Vals AI benchmarks · max

#7 / 38
75.1%[68.1% to 82.2%]$9.94On frontier. Evidence grade A of A to D: strongest.15 Nov 2026
FrontierCode

Epoch AI Benchmarking Hub · medium · Codex CLI

#7 / 51
50.2%UnknownStandardNot stated
LMArena WebDev

LMArena · Overall · max

#7 / 156
1,742[1,733 to 1,752]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev · max

#7 / 156
1,742[1,733 to 1,752]UnknownStandard15 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · max · Codex CLI

#8 / 145
44.3$5.88On frontier. Evidence grade A of A to D: strongest.30 Sept 2026
LMArena Search

LMArena · Overall · max

#8 / 42
1,230[1,223 to 1,238]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Industry mathematical · max

#8 / 402Tied #1
1,521[1,499 to 1,542]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Diagram · max

#8 / 123
1,327[1,307 to 1,347]UnknownStandard9 Nov 2026
LMArena WebDev

LMArena · Webdev react · max

#8 / 140
1,745[1,734 to 1,757]UnknownStandard15 Nov 2026
Vibe Code Bench

Vals AI benchmarks · max · OpenHands

#8 / 38
88.9%[85.2% to 92.6%]$6.23Standard15 Nov 2026
LMArena Agent

LMArena · Overall · max

#9 / 64
0.1[0.1 to 0.1]$0.99On frontier. Evidence grade A of A to D: strongest.15 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · max

#9 / 38
55.1%[51.5% to 58.6%]$1.72On frontier. Evidence grade A of A to D: strongest.15 Nov 2026
LMArena Vision (style control)

LMArena · Chinese · max

#10 / 135Tied #1
1,348[1,304 to 1,392]UnknownStandard9 Nov 2026
Vals Index

Vals AI benchmarks · max

#10 / 38
63.9%[61.8% to 66%]$3.61On frontier. Evidence grade A of A to D: strongest.15 Nov 2026
LMArena Agent

LMArena · Overall · medium

#12 / 64
0.1[0.1 to 0.1]$0.37On frontier. Evidence grade A of A to D: strongest.15 Nov 2026
LMArena Text (style control)

LMArena · Coding · max

#12 / 416
1,540[1,527 to 1,552]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry software and IT services · max

#12 / 421
1,527[1,518 to 1,537]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Overall · max

#12 / 167
1,293[1,282 to 1,303]UnknownStandard9 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · xhigh

#13 / 299
91.7%UnknownStandardNot stated
ARC-AGI-2

Epoch AI Benchmarking Hub · high

#13 / 299
91.7%UnknownStandardNot stated
Bug Hunt Bench

Bug Hunt Bench · xhigh · Codex CLI

#13 / 145
42.7$4.35On frontier. Evidence grade A of A to D: strongest.30 Sept 2026
LMArena Text (style control)

LMArena · Industry legal and government · max

#13 / 393Tied #1
1,504[1,485 to 1,523]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Korean · max

#13 / 289Tied #1
1,468[1,428 to 1,507]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · English · max

#13 / 167Tied #1
1,287[1,269 to 1,305]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts English · max

#15 / 419
1,515[1,503 to 1,526]UnknownStandard11 Nov 2026
EMB

Vals AI benchmarks · max

#16 / 38Tied #1
71.1%[65.7% to 76.5%]$1.14Standard15 Nov 2026
LMArena Vision (style control)

LMArena · Humor · max

#16 / 102Tied #1
1,283[1,238 to 1,327]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · OCR · max

#17 / 123
1,300[1,288 to 1,312]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Industry mathematical · max

#18 / 402
1,498[1,476 to 1,520]UnknownStandard11 Nov 2026
Finance Agent

Vals AI benchmarks · max

#19 / 38
61.8%[57.7% to 66%]$1.63Standard15 Nov 2026
LMArena Text (style control)

LMArena · Expert · max

#19 / 371
1,525[1,507 to 1,542]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Math · max

#19 / 404
1,499[1,477 to 1,522]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · English · max

#21 / 421
1,491[1,482 to 1,501]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Instruction following · max

#23 / 421
1,482[1,472 to 1,492]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Polish · max

#23 / 243Tied #1
1,493[1,451 to 1,535]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Creative writing vision · max

#23 / 106
1,292[1,260 to 1,324]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Chinese · max

#24 / 395
1,531[1,513 to 1,549]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry life and physical and social science · max

#24 / 419
1,501[1,487 to 1,514]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Diagram · max

#25 / 123
1,312[1,292 to 1,332]UnknownStandard9 Nov 2026
Legal Research

Vals AI benchmarks · max

#26 / 38
38.5%[31.8% to 45.1%]$3.38Standard15 Nov 2026
LMArena Text (style control)

LMArena · Exclude ties · max

#27 / 421
1,494[1,486 to 1,503]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts · max

#27 / 421
1,504[1,497 to 1,512]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · German · max

#28 / 320
1,484[1,444 to 1,524]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Korean · max

#29 / 289
1,442[1,402 to 1,481]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Multi turn · max

#29 / 419
1,486[1,471 to 1,500]UnknownStandard11 Nov 2026
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · max

#29 / 38
6.3%[2.7% to 9.8%]$3.54Standard15 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · high · Codex CLI

#30 / 145
36.5$3.26On frontier. Evidence grade A of A to D: strongest.30 Sept 2026
LMArena Text (style control)

LMArena · Industry business and management and financial operations · max

#30 / 414
1,480[1,467 to 1,494]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry writing and literature and language · max

#30 / 420
1,467[1,457 to 1,478]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Overall · max

#30 / 421
1,481[1,475 to 1,487]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Russian · max

#30 / 385
1,485[1,471 to 1,499]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Overall · max

#30 / 167
1,286[1,275 to 1,296]UnknownStandard9 Nov 2026
Tax Agent

Vals AI benchmarks · max

#30 / 38
21.8%[15.9% to 27.6%]$1.32Standard15 Nov 2026
LMArena Text (style control)

LMArena · Spanish · max

#31 / 304Tied #1
1,473[1,445 to 1,501]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Chinese · max

#31 / 135
1,340[1,297 to 1,384]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Expert · max

#32 / 371
1,498[1,480 to 1,515]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Non English · max

#32 / 421
1,468[1,461 to 1,476]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Longer query · max

#33 / 399
1,488[1,479 to 1,497]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · OCR · max

#33 / 123
1,291[1,279 to 1,303]UnknownStandard9 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · medium

#34 / 299
86.7%UnknownStandardNot stated
LMArena Text (style control)

LMArena · Industry entertainment and sports and media · max

#34 / 419
1,453[1,441 to 1,464]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Japanese · max

#36 / 288
1,462[1,423 to 1,500]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · French · max

#37 / 301
1,492[1,465 to 1,519]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · English · max

#37 / 167
1,280[1,263 to 1,298]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · Humor · max

#37 / 102
1,272[1,228 to 1,316]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Math · max

#39 / 404
1,475[1,453 to 1,498]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Instruction following · max

#41 / 421
1,463[1,453 to 1,473]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Creative writing · max

#43 / 419
1,449[1,436 to 1,461]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry legal and government · max

#44 / 393
1,472[1,453 to 1,491]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Coding · max

#46 / 416
1,486[1,474 to 1,498]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Japanese · max

#48 / 288
1,427[1,389 to 1,466]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Hard prompts English · max

#50 / 419
1,472[1,460 to 1,483]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Hard prompts · max

#52 / 421
1,466[1,458 to 1,473]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry medicine and healthcare · max

#52 / 391
1,483[1,462 to 1,503]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry software and IT services · max

#53 / 421
1,478[1,468 to 1,487]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Longer query · max

#53 / 399
1,459[1,450 to 1,468]UnknownStandard11 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · medium · Codex CLI

#54 / 145
29$1.76On frontier. Evidence grade A of A to D: strongest.30 Sept 2026
LMArena Text

LMArena · Chinese · max

#54 / 395
1,501[1,484 to 1,519]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Russian · max

#54 / 385
1,452[1,438 to 1,466]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry writing and literature and language · max

#56 / 420
1,440[1,430 to 1,451]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry business and management and financial operations · max

#57 / 414
1,441[1,428 to 1,455]UnknownStandard11 Nov 2026
LMArena Text

LMArena · German · max

#58 / 320
1,449[1,410 to 1,488]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Polish · max

#59 / 243
1,446[1,405 to 1,487]UnknownStandard11 Nov 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · low

#61 / 299
76.7%UnknownStandardNot stated
LMArena Text

LMArena · Non English · max

#62 / 421
1,435[1,428 to 1,443]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Exclude ties · max

#63 / 421
1,442[1,434 to 1,450]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry entertainment and sports and media · max

#64 / 419
1,421[1,410 to 1,433]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Overall · max

#66 / 421
1,445[1,439 to 1,451]UnknownStandard11 Nov 2026
LMArena Text

LMArena · English · max

#67 / 421
1,451[1,442 to 1,461]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry life and physical and social science · max

#70 / 419
1,457[1,444 to 1,470]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Multi turn · max

#70 / 419
1,449[1,435 to 1,463]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Creative writing · max

#72 / 419
1,420[1,407 to 1,432]UnknownStandard11 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · low · Codex CLI

#78 / 145
22.5$1.34On frontier. Evidence grade A of A to D: strongest.1 Oct 2026
LMArena Text

LMArena · Spanish · max

#85 / 304
1,439[1,411 to 1,466]UnknownStandard11 Nov 2026
LMArena Text

LMArena · French · max

#88 / 301
1,449[1,422 to 1,475]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry medicine and healthcare · max

#118 / 391
1,429[1,409 to 1,449]UnknownStandard11 Nov 2026

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.
    low

    76.7

    medium

    86.7

    high

    91.7

    xhigh

    91.7

    max

    94.2

  • Bug Hunt Bench

    Bug Hunt Bench · Codex CLI

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

    22.5

    $1.34

    medium

    29.0

    $1.76

    high

    36.5

    $3.26

    xhigh

    42.7

    $4.35

    max

    44.3

    $5.88

  • LMArena Agent

    LMArena

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

    0.1

    $0.37

    max

    0.1

    $0.99

Closest head-to-head rivals

Direct pairwise win rates for GPT-6.1 Sol across shared benchmark boards.