Acropolis

GLM-5.3

Z.aiGrade A. Evidence grade A of A to D: strongest.Open weightsReleased 14 Aug 2026Within-domain variance
Compare with peers

Domain profile

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

  • Agentic (14 boards): 57 (scale 0 to 100)Agentic (14 boards)57
  • Coding (9 boards): 53 (scale 0 to 100)Coding (9 boards)53
  • Finance (5 boards): 68 (scale 0 to 100)Finance (5 boards)68
  • General text (2 boards): 94 (scale 0 to 100)General text (2 boards)94
  • Legal (3 boards): 70 (scale 0 to 100)Legal (3 boards)70
  • Science (1 boards): 83 (scale 0 to 100)Science (1 boards)83

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
LMArena Text (style control)

LMArena · German · max

#7 / 320
1,510[1,483 to 1,537]UnknownStandard11 Nov 2026
Legal Research

Vals AI benchmarks · max

#7 / 38Tied #1
49%[42.2% to 55.8%]$2.24Standard15 Nov 2026
Tax Agent

Vals AI benchmarks · max

#7 / 38Tied #1
43.3%[36.3% to 50.3%]$1.80Standard15 Nov 2026
LMArena Text (style control)

LMArena · French · max

#9 / 301
1,513[1,493 to 1,532]UnknownStandard11 Nov 2026
LMArena Text

LMArena · German · max

#12 / 320
1,497[1,470 to 1,524]UnknownStandard11 Nov 2026
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · max

#12 / 38
14.6%[9% to 20.2%]$3.99Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · max

#13 / 38Tied #1
65%[61.1% to 68.9%]$1.03Standard15 Nov 2026
LMArena Text

LMArena · Chinese · max

#15 / 395
1,535[1,522 to 1,547]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Expert · max

#15 / 371
1,519[1,507 to 1,530]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry mathematical · max

#15 / 402
1,502[1,487 to 1,517]UnknownStandard11 Nov 2026
LMArena Text

LMArena · French · max

#16 / 301
1,498[1,479 to 1,517]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Math · max

#16 / 404
1,500[1,483 to 1,516]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Chinese · max

#16 / 395
1,540[1,527 to 1,553]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Math · max

#16 / 404
1,501[1,485 to 1,518]UnknownStandard11 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · max

#16 / 38
38.9%[35.3% to 42.5%]$9.37Standard15 Nov 2026
Vals Index

Vals AI benchmarks · max

#16 / 38
59.8%[57.7% to 62%]$8.52Standard15 Nov 2026
Vending-Bench 2

Epoch AI Benchmarking Hub · Default

#16 / 72
$8,163.61UnknownStandardNot stated
Code Migration

Vals AI benchmarks · Default

#17 / 38
54.1%[47.8% to 60.4%]$33.52Standard15 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts English · max

#17 / 419
1,514[1,507 to 1,522]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Hard prompts English · max

#18 / 419
1,500[1,492 to 1,507]UnknownStandard11 Nov 2026
LMArena Text

LMArena · English · max

#19 / 421
1,483[1,477 to 1,489]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Hard prompts · max

#20 / 421
1,493[1,487 to 1,498]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry life and physical and social science · max

#20 / 419
1,499[1,490 to 1,508]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Instruction following · max

#20 / 421
1,483[1,476 to 1,490]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Expert · max

#20 / 371
1,522[1,511 to 1,534]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry life and physical and social science · max

#20 / 419
1,508[1,499 to 1,517]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry mathematical · max

#20 / 402
1,502[1,487 to 1,517]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Longer query · max

#21 / 399
1,486[1,480 to 1,493]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Japanese · max

#22 / 288
1,480[1,453 to 1,507]UnknownStandard11 Nov 2026
EMB

Vals AI benchmarks · max

#23 / 38
69.1%[63.9% to 74.3%]$3.79Standard15 Nov 2026
LMArena Text (style control)

LMArena · English · max

#23 / 421
1,491[1,485 to 1,497]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts · max

#23 / 421
1,507[1,502 to 1,513]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry medicine and healthcare · max

#24 / 391
1,500[1,487 to 1,513]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Instruction following · max

#24 / 421
1,481[1,474 to 1,487]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry writing and literature and language · max

#25 / 420
1,468[1,461 to 1,476]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry software and IT services · max

#26 / 421
1,495[1,489 to 1,502]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Japanese · max

#26 / 288
1,470[1,442 to 1,497]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Overall · max

#26 / 421
1,475[1,470 to 1,480]UnknownStandard11 Nov 2026
FrontierCode

Epoch AI Benchmarking Hub · max · devin-cli

#27 / 51
40.1%UnknownStandardNot stated
LMArena Text

LMArena · Creative writing · max

#27 / 419
1,459[1,451 to 1,467]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Exclude ties · max

#27 / 421
1,481[1,475 to 1,487]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry medicine and healthcare · max

#27 / 391
1,483[1,469 to 1,496]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry software and IT services · max

#27 / 421
1,515[1,508 to 1,521]UnknownStandard11 Nov 2026
Vibe Code Bench

Vals AI benchmarks · max · OpenHands

#27 / 38
78.1%[70.4% to 85.8%]$12.45Standard15 Nov 2026
LMArena Text

LMArena · Coding · max

#28 / 416
1,499[1,491 to 1,506]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Overall · max

#28 / 421
1,482[1,477 to 1,486]UnknownStandard11 Nov 2026
LMArena WebDev

LMArena · Webdev react · max

#28 / 140
1,641[1,633 to 1,649]UnknownStandard15 Nov 2026
LMArena Text (style control)

LMArena · Exclude ties · max

#30 / 421
1,491[1,485 to 1,497]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Korean · max

#30 / 289
1,443[1,416 to 1,470]UnknownStandard11 Nov 2026
LMArena WebDev

LMArena · Overall · max

#30 / 156
1,622[1,615 to 1,629]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev · max

#30 / 156
1,622[1,615 to 1,629]UnknownStandard15 Nov 2026
LMArena Text

LMArena · Korean · max

#31 / 289
1,440[1,413 to 1,467]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry writing and literature and language · max

#31 / 420
1,467[1,460 to 1,474]UnknownStandard11 Nov 2026
LMArena Agent

LMArena · Overall · max

#32 / 64
0[-0 to 0]$0.86Standard15 Nov 2026
LMArena Text

LMArena · Non English · max

#32 / 421
1,462[1,456 to 1,467]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Longer query · max

#32 / 399
1,488[1,482 to 1,495]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry legal and government · max

#33 / 393
1,481[1,468 to 1,493]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Multi turn · max

#33 / 419
1,474[1,465 to 1,484]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Coding · max

#33 / 416
1,522[1,514 to 1,530]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Creative writing · max

#33 / 419
1,456[1,448 to 1,464]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry entertainment and sports and media · max

#34 / 419
1,447[1,440 to 1,455]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Russian · max

#34 / 385
1,472[1,463 to 1,482]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Spanish · max

#34 / 304
1,471[1,452 to 1,490]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry business and management and financial operations · max

#35 / 414
1,461[1,453 to 1,470]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Spanish · max

#35 / 304
1,467[1,448 to 1,486]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry business and management and financial operations · max

#35 / 414
1,476[1,467 to 1,484]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Non English · max

#35 / 421
1,467[1,462 to 1,473]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Polish · max

#36 / 243
1,463[1,436 to 1,490]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Russian · max

#36 / 385
1,480[1,470 to 1,489]UnknownStandard11 Nov 2026
LMArena WebDev

LMArena · Webdev HTML · max

#36 / 155
1,597[1,581 to 1,612]UnknownStandard15 Nov 2026
LMArena Text (style control)

LMArena · Industry entertainment and sports and media · max

#37 / 419
1,448[1,441 to 1,456]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Multi turn · max

#37 / 419
1,482[1,473 to 1,492]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry legal and government · max

#39 / 393
1,482[1,469 to 1,494]UnknownStandard11 Nov 2026
CursorBench

CursorBench · max · Cursor

#44 / 91
42.6%$5.05StandardNot stated
GPQA diamond

Epoch AI Benchmarking Hub · max

#44 / 338
90.9%[87.8% to 94%]UnknownStandard24 Aug 2026
LMArena Text (style control)

LMArena · Polish · max

#44 / 243
1,475[1,448 to 1,502]UnknownStandard11 Nov 2026
CursorBench

CursorBench · high · Cursor

#55 / 91
38%$3.24StandardNot stated
CursorBench

CursorBench · low · Cursor

#69 / 91
33.3%$2.04StandardNot stated
Bug Hunt Bench

Bug Hunt Bench · max · Claude Code

#91 / 145
19$15.93Standard10 Sept 2026

Reasoning effort scaling

Same benchmark and harness measured across multiple reasoning effort settings.

  • CursorBench

    CursorBench · Cursor

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

    33.3

    $2.04

    high

    38.0

    $3.24

    max

    42.6

    $5.05

Closest head-to-head rivals

Direct pairwise win rates for GLM-5.3 across shared benchmark boards.