Acropolis

GLM-5.3 Flash

Z.aiGrade A. Evidence grade A of A to D: strongest.Open weightsReleased 20 Aug 2026Within-domain variance
Compare with peers

Domain profile

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

  • Agentic (13 boards): 33 (scale 0 to 100)Agentic (13 boards)33
  • Coding (9 boards): 33 (scale 0 to 100)Coding (9 boards)33
  • Finance (4 boards): 31 (scale 0 to 100)Finance (4 boards)31
  • General text (3 boards): 92 (scale 0 to 100)General text (3 boards)92
  • Legal (3 boards): 39 (scale 0 to 100)Legal (3 boards)39
  • Science (2 boards): 75 (scale 0 to 100)Science (2 boards)75
  • Vision (2 boards): 85 (scale 0 to 100)Vision (2 boards)85

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
LMArena Vision

LMArena · Entity recognition · Default

#10 / 52Tied #1
1,279[1,203 to 1,354]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Math · Default

#11 / 404
1,505[1,492 to 1,517]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Entity recognition · Default

#13 / 52Tied #1
1,236[1,161 to 1,311]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Math · Default

#15 / 404
1,503[1,491 to 1,516]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · English · Default

#15 / 167
1,299[1,291 to 1,307]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Coding · Default

#16 / 416
1,509[1,503 to 1,516]UnknownStandard11 Nov 2026
Legal Research

Vals AI benchmarks · max

#16 / 38
45.2%[38.4% to 52%]$0.24Standard15 Nov 2026
LMArena Vision

LMArena · Diagram · Default

#17 / 123
1,321[1,313 to 1,329]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Industry software and IT services · Default

#18 / 421
1,501[1,495 to 1,507]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Chinese · Default

#18 / 135
1,359[1,341 to 1,376]UnknownStandard9 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · max

#18 / 38
37.4%[32.1% to 42.6%]$1.22Standard15 Nov 2026
LMArena Vision

LMArena · Humor · Default

#19 / 102
1,291[1,273 to 1,308]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Expert · Default

#20 / 371
1,511[1,502 to 1,521]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Homework · Default

#20 / 118
1,323[1,313 to 1,332]UnknownStandard9 Nov 2026
Code Migration

Vals AI benchmarks · Default

#21 / 38
50.6%[42.8% to 58.5%]$8.27Standard15 Nov 2026
LMArena Text

LMArena · Hard prompts · Default

#22 / 421
1,490[1,485 to 1,496]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Instruction following · Default

#22 / 421
1,480[1,474 to 1,486]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Chinese · Default

#22 / 135
1,334[1,316 to 1,351]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · Homework · Default

#22 / 118
1,312[1,302 to 1,321]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Hard prompts English · Default

#23 / 419
1,492[1,485 to 1,498]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Spanish · Default

#23 / 304
1,476[1,462 to 1,491]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Humor · Default

#23 / 102
1,268[1,251 to 1,286]UnknownStandard9 Nov 2026
LMArena WebDev

LMArena · Image to webdev · Default

#23 / 64
1,582[1,574 to 1,590]UnknownStandard15 Nov 2026
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · max

#23 / 38
10.4%[5.4% to 15.5%]$0.32Standard15 Nov 2026
Tax Agent

Vals AI benchmarks · max

#23 / 38
31.6%[25% to 38.1%]$0.14Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · max

#24 / 38
60.6%[56.6% to 64.6%]$0.07Standard15 Nov 2026
LMArena Vision

LMArena · OCR · Default

#24 / 123
1,303[1,298 to 1,309]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · Overall · Default

#24 / 167
1,295[1,289 to 1,300]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Chinese · Default

#25 / 395
1,528[1,518 to 1,538]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Spanish · Default

#25 / 304
1,475[1,460 to 1,490]UnknownStandard11 Nov 2026
LMArena Text

LMArena · French · Default

#26 / 301
1,489[1,474 to 1,504]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Chinese · Default

#26 / 395
1,529[1,518 to 1,539]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry software and IT services · Default

#26 / 421
1,515[1,509 to 1,521]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Diagram · Default

#27 / 123
1,310[1,302 to 1,318]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Industry mathematical · Default

#28 / 402
1,491[1,479 to 1,503]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Longer query · Default

#28 / 399
1,481[1,476 to 1,487]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Coding · Default

#28 / 416
1,526[1,520 to 1,533]UnknownStandard11 Nov 2026
EMB

Vals AI benchmarks · max

#29 / 38
66.3%[61% to 71.5%]$1.43Standard15 Nov 2026
LMArena Text

LMArena · Industry life and physical and social science · Default

#29 / 419
1,485[1,477 to 1,492]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · French · Default

#29 / 301
1,499[1,484 to 1,514]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry medicine and healthcare · Default

#30 / 391
1,480[1,469 to 1,491]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Overall · Default

#30 / 421
1,472[1,467 to 1,476]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry medicine and healthcare · Default

#30 / 391
1,493[1,482 to 1,504]UnknownStandard11 Nov 2026
LMArena Text

LMArena · English · Default

#31 / 421
1,477[1,471 to 1,482]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Exclude ties · Default

#31 / 421
1,477[1,472 to 1,483]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Expert · Default

#31 / 371
1,511[1,502 to 1,520]UnknownStandard11 Nov 2026
FrontierCode

Epoch AI Benchmarking Hub · max · devin-cli

#32 / 51
31.8%UnknownStandardNot stated
Vals Index

Vals AI benchmarks · max

#32 / 38
50.9%[48.4% to 53.3%]$2.05Standard15 Nov 2026
LMArena Text

LMArena · Japanese · Default

#33 / 288
1,453[1,432 to 1,475]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Non English · Default

#33 / 421
1,462[1,457 to 1,467]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · English · Default

#33 / 167
1,274[1,266 to 1,282]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts English · Default

#34 / 419
1,502[1,495 to 1,508]UnknownStandard11 Nov 2026
LMArena WebDev

LMArena · Webdev react · Default

#34 / 140
1,628[1,622 to 1,635]UnknownStandard15 Nov 2026
LMArena Text (style control)

LMArena · Hard prompts · Default

#35 / 421
1,500[1,495 to 1,506]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry mathematical · Default

#35 / 402
1,490[1,478 to 1,502]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Instruction following · Default

#35 / 421
1,474[1,468 to 1,480]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · OCR · Default

#35 / 123
1,287[1,281 to 1,292]UnknownStandard9 Nov 2026
Vibe Code Bench

Vals AI benchmarks · max · OpenHands

#35 / 38
30.8%[20.4% to 41.1%]$4.70Standard15 Nov 2026
LMArena Text

LMArena · Industry business and management and financial operations · Default

#36 / 414
1,461[1,454 to 1,469]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Overall · Default

#36 / 421
1,475[1,471 to 1,480]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Overall · Default

#36 / 167
1,272[1,266 to 1,277]UnknownStandard9 Nov 2026
LMArena WebDev

LMArena · Overall · Default

#36 / 156
1,609[1,603 to 1,615]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev · Default

#36 / 156
1,609[1,603 to 1,615]UnknownStandard15 Nov 2026
LMArena Text

LMArena · Industry legal and government · Default

#37 / 393
1,478[1,468 to 1,489]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry writing and literature and language · Default

#37 / 420
1,457[1,451 to 1,463]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Russian · Default

#37 / 385
1,467[1,459 to 1,475]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Japanese · Default

#37 / 288
1,460[1,439 to 1,482]UnknownStandard11 Nov 2026
LMArena Vision

LMArena · Creative writing vision · Default

#37 / 106
1,272[1,259 to 1,284]UnknownStandard9 Nov 2026
LMArena Text

LMArena · Multi turn · Default

#38 / 419
1,471[1,464 to 1,479]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Non English · Default

#38 / 421
1,464[1,459 to 1,469]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Industry entertainment and sports and media · Default

#39 / 419
1,442[1,435 to 1,448]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Korean · Default

#39 / 289
1,427[1,406 to 1,449]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · English · Default

#40 / 421
1,481[1,476 to 1,487]UnknownStandard11 Nov 2026
LMArena WebDev

LMArena · Webdev HTML · Default

#40 / 155
1,576[1,565 to 1,588]UnknownStandard15 Nov 2026
LMArena Text (style control)

LMArena · Industry life and physical and social science · Default

#41 / 419
1,491[1,483 to 1,498]UnknownStandard11 Nov 2026
LMArena Agent

LMArena · Overall · Default

#44 / 64
-0[-0 to -0]$0.14Standard15 Nov 2026
LMArena Text

LMArena · Creative writing · Default

#44 / 419
1,442[1,435 to 1,449]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry business and management and financial operations · Default

#44 / 414
1,471[1,464 to 1,478]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Exclude ties · Default

#45 / 421
1,484[1,478 to 1,489]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Korean · Default

#45 / 289
1,430[1,409 to 1,452]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Russian · Default

#47 / 385
1,472[1,464 to 1,480]UnknownStandard11 Nov 2026
LMArena Vision (style control)

LMArena · Creative writing vision · Default

#50 / 106
1,246[1,233 to 1,258]UnknownStandard9 Nov 2026
LMArena Text (style control)

LMArena · Multi turn · Default

#51 / 419
1,476[1,469 to 1,484]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry legal and government · Default

#52 / 393
1,475[1,465 to 1,486]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Longer query · Default

#52 / 399
1,480[1,474 to 1,485]UnknownStandard11 Nov 2026
GPQA diamond

Epoch AI Benchmarking Hub · max

#54 / 338
90.2%[86.8% to 93.5%]UnknownStandard26 Aug 2026
LMArena Text (style control)

LMArena · Industry writing and literature and language · Default

#55 / 420
1,452[1,446 to 1,459]UnknownStandard11 Nov 2026
LMArena Text

LMArena · German · Default

#56 / 320
1,452[1,431 to 1,474]UnknownStandard11 Nov 2026
LMArena Text

LMArena · Polish · Default

#56 / 243
1,448[1,428 to 1,468]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Industry entertainment and sports and media · Default

#56 / 419
1,439[1,432 to 1,445]UnknownStandard11 Nov 2026
CursorBench

CursorBench · max · Cursor

#59 / 91
36.8%$0.39StandardNot stated
LMArena Text (style control)

LMArena · German · Default

#64 / 320
1,459[1,438 to 1,481]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Creative writing · Default

#67 / 419
1,435[1,428 to 1,442]UnknownStandard11 Nov 2026
LMArena Text (style control)

LMArena · Polish · Default

#72 / 243
1,455[1,435 to 1,476]UnknownStandard11 Nov 2026
CursorBench

CursorBench · high · Cursor

#74 / 91
31.1%$0.25StandardNot stated
CursorBench

CursorBench · low · Cursor

#85 / 91
26.9%$0.15StandardNot stated
ARC-AGI-2

Epoch AI Benchmarking Hub · max

#89 / 299
65.8%UnknownStandardNot stated
Bug Hunt Bench

Bug Hunt Bench · max · Claude Code

#101 / 145
17.7$1.03Standard15 Sept 2026
Bug Hunt Bench

Bug Hunt Bench · high · Claude Code

#104 / 145
16$0.94Standard13 Sept 2026
Bug Hunt Bench

Bug Hunt Bench · Claude Code

#118 / 145
13$0.79Standard27 Aug 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · high

#126 / 299
50.1%UnknownStandardNot stated
Bug Hunt Bench

Bug Hunt Bench · low · Claude Code

#129 / 145
9$0.42Standard13 Sept 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · low

#155 / 299
27.9%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.
    low

    27.9

    high

    50.1

    max

    65.8

  • Bug Hunt Bench

    Bug Hunt Bench · Claude Code

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

    9.0

    $0.42

    high

    16.0

    $0.94

  • CursorBench

    CursorBench · Cursor

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

    26.9

    $0.15

    high

    31.1

    $0.25

    max

    36.8

    $0.39

Closest head-to-head rivals

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