Benchize

Professional work benchmarks

Evaluate legal, financial, healthcare, and other specialist work. Check each task's domain, required expertise, and scoring method when interpreting model performance.

Published

HLAB-Vals-1.1 / criteria pass rate immigration

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax96.18%
  2. 2Gemini 4 ArgonHigh94.55%
  3. 3Claude Sonnet 5.5Max94%
View details →

HLAB-Vals-1.1 / criteria pass rate insurance

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh96.32%
  2. 2Muse WonderMax96.14%
  3. 3DeepSeek V4.1 Flash 1017Max94.49%
View details →

HLAB-Vals-1.1 / criteria pass rate intellectual property

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh98.4%
  2. 2Muse WonderMax97.15%
  3. 3Claude Sonnet 5.5Max96.62%
View details →

HLAB-Vals-1.1 / criteria pass rate international trade sanctions

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax96.41%
  2. 2Gemini 4 ArgonHigh96.22%
  3. 3DeepSeek V4.1 Flash 1017Max94.82%
View details →

HLAB-Vals-1.1 / criteria pass rate litigation dispute resolution

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh96.24%
  2. 2Muse WonderMax95.11%
  3. 3Gemini 3.8 FlashHigh94.17%
View details →

HLAB-Vals-1.1 / criteria pass rate real estate

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax97.23%
  2. 2Gemini 4 ArgonHigh97.09%
  3. 3GPT-6.1 SolMax96.21%
View details →

HLAB-Vals-1.1 / criteria pass rate structured finance securitization

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax94.19%
  2. 2Gemini 4 ArgonHigh94.03%
  3. 3Claude Sonnet 5.5Max93.55%
View details →

HLAB-Vals-1.1 / criteria pass rate tax

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh94.82%
  2. 2Muse WonderMax92.48%
  3. 3MiMo-V2.6-Pro91.47%
View details →

HLAB-Vals-1.1 / criteria pass rate trusts estates private client

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax96.18%
  2. 2Gemini 4 ArgonHigh95.99%
  3. 3DeepSeek V4.1 Flash 1017Max95.42%
View details →

HLAB-Vals-1.1 / criteria pass rate white collar defense investigations

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax96.97%
  2. 2Gemini 4 ArgonHigh95.46%
  3. 3Gemini 3.8 FlashHigh91.92%
View details →

HLAB-Vals-1.1 / data privacy cybersecurity

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh50%
  2. 2MiMo-V2.6-Pro40%
  3. 3Claude Opus 5.5Max30%
View details →

HLAB-Vals-1.1 / emerging companies venture capital

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax40%
  2. 2Gemini 3.8 FlashHigh30%
  3. 3GPT-6 AstraMax20%
View details →

HLAB-Vals-1.1 / employment labor

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax0%
  2. 2Claude Sonnet 5.5Max0%
  3. 3Gemini 3.8 FlashHigh0%
View details →

HLAB-Vals-1.1 / energy natural resources

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax70%
  2. 2Claude Opus 5.5Max50%
  3. 3DeepSeek V4 Pro 0813Max40%
View details →

HLAB-Vals-1.1 / environmental esg

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax20%
  2. 2Gemini 4 ArgonHigh20%
  3. 3GPT-6.1 SolMax0%
View details →

HLAB-Vals-1.1 / funds asset management

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh60%
  2. 2DeepSeek V4.1 Flash 1017Max40%
  3. 3MiMo-V2.6-Flash30%
View details →

HLAB-Vals-1.1 / healthcare life sciences

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh60%
  2. 2MiMo-V2.6-Pro40%
  3. 3DeepSeek V4.1 Flash 1017Max30%
View details →

HLAB-Vals-1.1 / immigration

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax40%
  2. 2DeepSeek V4 Pro 0813Max20%
  3. 3Claude Sonnet 5.5Max20%
View details →

HLAB-Vals-1.1 / insurance

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax30%
  2. 2Gemini 4 ArgonHigh20%
  3. 3GPT-6 AstraMax10%
View details →

HLAB-Vals-1.1 / intellectual property

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh40%
  2. 2Claude Sonnet 5.5Max30%
  3. 3Gemini 3.8 FlashHigh30%
View details →

HLAB-Vals-1.1 / international trade sanctions

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax30%
  2. 2Gemini 3.8 FlashHigh20%
  3. 3Gemini 4 ArgonHigh10%
View details →

HLAB-Vals-1.1 / litigation dispute resolution

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1DeepSeek V4.1 Flash 1017Max20%
  2. 2DeepSeek V4 Pro 0813Max20%
  3. 3MiMo-V2.6-Flash20%
View details →

HLAB-Vals-1.1 / overall

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh27.08%
  2. 2Muse WonderMax23.75%
  3. 3DeepSeek V4.1 Flash 1017Max14.17%
View details →

HLAB-Vals-1.1 / real estate

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1GPT-6.1 SolMax30%
  2. 2Muse WonderMax30%
  3. 3Claude Haiku 5.5Max20%
View details →

HLAB-Vals-1.1 / structured finance securitization

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh30%
  2. 2Claude Sonnet 5.5Max30%
  3. 3DeepSeek V4.1 Flash 1017Max20%
View details →

HLAB-Vals-1.1 / tax

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh20%
  2. 2MiMo-V2.6-Flash10%
  3. 3Gemini 3.8 FlashHigh10%
View details →

HLAB-Vals-1.1 / trusts estates private client

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Gemini 4 ArgonHigh30%
  2. 2Gemini 3.8 FlashHigh30%
  3. 3MiMo-V2.6-Pro30%
View details →

HLAB-Vals-1.1 / white collar defense investigations

Updated:

Models tested: 14

Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools

Top models · Vals AI

Latest result per model · 1.1

  1. 1Muse WonderMax20%
  2. 2GPT-6.1 SolMax0%
  3. 3Gemini 3.8 FlashHigh0%
View details →

LegalBench / conclusion tasks

Updated:

Models tested: 44

Evaluating language models on a wide range of open source legal reasoning tasks.

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5.1Max92.51%
  2. 2Claude Fable 5Max91.78%
  3. 3Claude Opus 5Max91.56%
View details →

LegalBench / interpretation tasks

Updated:

Models tested: 44

Evaluating language models on a wide range of open source legal reasoning tasks.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.5 FlashHigh88.68%
  2. 2Claude Fable 5.1Max86.9%
  3. 3Claude Fable 5Max86.86%
View details →

LegalBench / issue tasks

Updated:

Models tested: 44

Evaluating language models on a wide range of open source legal reasoning tasks.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 4 ArgonHigh92.94%
  2. 2Claude Fable 5.1Max92.72%
  3. 3Gemini 3.7 FlashHigh92.15%
View details →

LegalBench / overall

Updated:

Models tested: 44

Evaluating language models on a wide range of open source legal reasoning tasks.

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5Max88.56%
  2. 2Claude Fable 5.1Max88.51%
  3. 3Gemini 4 ArgonHigh88.3%
View details →

LegalBench / rhetoric tasks

Updated:

Models tested: 44

Evaluating language models on a wide range of open source legal reasoning tasks.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.1 Pro PreviewHigh84.14%
  2. 2Gemini 3 Flash PreviewHigh83.87%
  3. 3Claude Fable 5Max83.77%
View details →

LegalBench / rule tasks

Updated:

Models tested: 44

Evaluating language models on a wide range of open source legal reasoning tasks.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 4 ArgonHigh90.77%
  2. 2Claude Fable 5Max88.33%
  3. 3Gemini 3.1 Pro PreviewHigh88.29%
View details →

PRBench Legal

Updated:

Models tested: 26

Realistic legal reasoning tasks graded against detailed expert-written criteria for professional work.

Top models · scale

Latest result per model · Version not reported

  1. 1Muse Spark 1.361.56 source score
  2. 2Muse Spark Open59.76 source score
  3. 3Muse Spark 1.157.05 source score
View details →

TaxEval v2 / correctness

Updated:

Models tested: 40

A Vals-created set of questions and responses to tax questions

Top models · Vals AI

Latest result per model · 2

  1. 1Muse Spark 1.2Extra high68.03%
  2. 2Muse Spark 1.1Extra high66.89%
  3. 3GPT-5.6 LunaMax66.56%
View details →

TaxEval v2 / overall

Updated:

Models tested: 40

A Vals-created set of questions and responses to tax questions

Top models · Vals AI

Latest result per model · 2

  1. 1Muse Spark 1.2Extra high80.38%
  2. 2Muse Spark 1.1Extra high79.72%
  3. 3Claude Sonnet 4.6Max77.11%
View details →

TaxEval v2 / stepwise reasoning

Updated:

Models tested: 40

A Vals-created set of questions and responses to tax questions

Top models · Vals AI

Latest result per model · 2

  1. 1Muse Spark 1.2Extra high92.72%
  2. 2Muse Spark 1.1Extra high92.56%
  3. 3Claude Sonnet 4.6Max88.39%
View details →

CorpFin v2 / exact pages

Updated:

Models tested: 39

A private benchmark evaluating understanding of long-context credit agreements

Top models · Vals AI

Latest result per model · 2

  1. 1Claude Opus 568.65%
  2. 2Gemini 3 Flash PreviewHigh68.07%
  3. 3Claude Fable 5Max66.9%
View details →

CorpFin v2 / max fitting context

Updated:

Models tested: 39

A private benchmark evaluating understanding of long-context credit agreements

Top models · Vals AI

Latest result per model · 2

  1. 1Claude Fable 5Max77.04%
  2. 2Claude Opus 575.99%
  3. 3Muse Spark 1.1Extra high74.36%
View details →

CorpFin v2 / overall

Updated:

Models tested: 39

A private benchmark evaluating understanding of long-context credit agreements

Top models · Vals AI

Latest result per model · 2

  1. 1Claude Opus 573.19%
  2. 2Claude Fable 5Max71.83%
  3. 3Muse Spark 1.1Extra high71.29%
View details →

CorpFin v2 / shared max context

Updated:

Models tested: 39

A private benchmark evaluating understanding of long-context credit agreements

Top models · Vals AI

Latest result per model · 2

  1. 1Claude Opus 574.94%
  2. 2Muse Spark 1.1Extra high73.08%
  3. 3Muse Spark 1.2Extra high72.73%
View details →

MortgageTax / numerical extraction

Updated:

Models tested: 31

Evaluating reading and understanding tax certificates as images

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Opus 581.4%
  2. 2Claude Opus 4.8Max79.73%
  3. 3Claude Opus 4.7Max78.7%
View details →

MortgageTax / overall

Updated:

Models tested: 31

Evaluating reading and understanding tax certificates as images

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Opus 572.06%
  2. 2Claude Fable 5.1Max70.79%
  3. 3Claude Opus 4.7Max70.27%
View details →

MortgageTax / semantic extraction

Updated:

Models tested: 31

Evaluating reading and understanding tax certificates as images

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5.1Max63.99%
  2. 2Claude Opus 562.72%
  3. 3Gemini 3.1 Pro PreviewHigh62.4%
View details →

ClaimBench / direct support

Conquered

Updated:

Models tested: 5

Does a 10-K excerpt support a claim, contradict it, or say nothing about it?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Sonnet 5.5Max100%
  2. 2GPT-6 LunaMax100%
  3. 3Claude Haiku 5.5Max100%
View details →

ClaimBench / missing evidence

Conquered

Updated:

Models tested: 5

Does a 10-K excerpt support a claim, contradict it, or say nothing about it?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Haiku 5.5Max100%
  2. 2GPT-6 LunaMax100%
  3. 3Claude Opus 5.5Max100%
View details →

ClaimBench / non supporting citation

Conquered

Updated:

Models tested: 5

Does a 10-K excerpt support a claim, contradict it, or say nothing about it?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Opus 5.5Max100%
  2. 2GPT-6.1 SolMax100%
  3. 3GPT-6 LunaMax100%
View details →

ClaimBench / numeric discrepancy

Conquered

Updated:

Models tested: 5

Does a 10-K excerpt support a claim, contradict it, or say nothing about it?

Top models · Vals AI

Latest result per model · 1

  1. 1GPT-6.1 SolMax100%
  2. 2GPT-6 LunaMax100%
  3. 3Claude Opus 5.5Max100%
View details →

ClaimBench / overall

Conquered

Updated:

Models tested: 5

Does a 10-K excerpt support a claim, contradict it, or say nothing about it?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Opus 5.5Max100%
  2. 2Claude Sonnet 5.5Max100%
  3. 3GPT-6.1 SolMax100%
View details →

ClaimBench / subtle contradiction

Conquered

Updated:

Models tested: 5

Does a 10-K excerpt support a claim, contradict it, or say nothing about it?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Haiku 5.5Max100%
  2. 2Claude Sonnet 5.5Max100%
  3. 3Claude Opus 5.5Max100%
View details →

PRBench Finance

Updated:

Models tested: 22

Realistic financial reasoning and decision tasks graded against detailed expert-written criteria.

Top models · scale

Latest result per model · Version not reported

  1. 1Muse Spark 1.359.54 source score
  2. 2Muse Spark 1.155.01 source score
  3. 3Claude Fable 553.86 source score
View details →

Public Benefits Bench v1 / both

Updated:

Models tested: 8

Can AI help people navigate SNAP benefits?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5Max71.65%
  2. 2Claude Opus 4.8Max62.11%
  3. 3MiniMax M360.69%
View details →

Public Benefits Bench v1 / multi turn

Updated:

Models tested: 8

Can AI help people navigate SNAP benefits?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5Max65.22%
  2. 2Claude Opus 4.8Max58.46%
  3. 3Gemini 3.5 FlashHigh53.99%
View details →

Public Benefits Bench v1 / neither

Updated:

Models tested: 8

Can AI help people navigate SNAP benefits?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5Max63.4%
  2. 2Claude Opus 4.8Max53.11%
  3. 3Gemini 3.5 FlashHigh50.27%
View details →

Public Benefits Bench v1 / overall

Updated:

Models tested: 8

Can AI help people navigate SNAP benefits?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5Max71.65%
  2. 2Claude Opus 4.8Max62.11%
  3. 3MiniMax M360.69%
View details →

Public Benefits Bench v1 / web search

Updated:

Models tested: 8

Can AI help people navigate SNAP benefits?

Top models · Vals AI

Latest result per model · 1

  1. 1Claude Fable 5Max67.73%
  2. 2Claude Opus 4.8Max58.93%
  3. 3Gemini 3.5 FlashHigh54.13%
View details →

CaseLaw v2 / overall

Updated:

Models tested: 17

Private question-answer benchmark over Canadian court-cases.

Top models · Vals AI

Latest result per model · 2

  1. 1Claude Opus 4.768.38%
  2. 2GPT-5.5Extra high66.24%
  3. 3Gemini 3.1 Pro PreviewHigh64.85%
View details →

MedQA / asian

Updated:

Models tested: 12

Evaluating language model bias in medical questions.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.1 Pro PreviewHigh96.3%
  2. 2Gemini 3 Flash PreviewHigh95.85%
  3. 3MiniMax M2.592.6%
View details →

MedQA / black

Updated:

Models tested: 12

Evaluating language model bias in medical questions.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.1 Pro PreviewHigh96.8%
  2. 2Gemini 3 Flash PreviewHigh95.7%
  3. 3Claude Sonnet 4.6Thinking92.75%
View details →

MedQA / hispanic

Updated:

Models tested: 12

Evaluating language model bias in medical questions.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.1 Pro PreviewHigh95.95%
  2. 2Gemini 3 Flash PreviewHigh95.4%
  3. 3MiniMax M2.592.3%
View details →

MedQA / indigenous

Updated:

Models tested: 12

Evaluating language model bias in medical questions.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.1 Pro PreviewHigh96.1%
  2. 2Gemini 3 Flash PreviewHigh95.6%
  3. 3MiniMax M2.592.6%
View details →

MedQA / overall

Updated:

Models tested: 12

Evaluating language model bias in medical questions.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.1 Pro PreviewHigh96.37%
  2. 2Gemini 3 Flash PreviewHigh95.81%
  3. 3MiniMax M2.592.53%
View details →

MedQA / unbiased

Updated:

Models tested: 12

Evaluating language model bias in medical questions.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3.1 Pro PreviewHigh96.8%
  2. 2Gemini 3 Flash PreviewHigh96%
  3. 3MiniMax M2.593.2%
View details →

MedQA / white

Updated:

Models tested: 12

Evaluating language model bias in medical questions.

Top models · Vals AI

Latest result per model · 1

  1. 1Gemini 3 Flash PreviewHigh96.3%
  2. 2Gemini 3.1 Pro PreviewHigh96.25%
  3. 3MiniMax M2.592.35%
View details →

TutorBench

Updated:

Models tested: 5

Large Language Models serve as on-demand tutors for learners worldwide, yet, a critical evaluation gap exists. While most benchmarks assess an LLM's ability to solve problems,…

Top models · scale

Latest result per model · Version not reported

  1. 1Gemini 2.5 Pro Preview 06-0555.65 source score
  2. 2Claude Opus 4.6Non-thinking53.55 source score
  3. 3Gemini 3.1 Pro Preview52.99 source score
View details →

Arena text / industry business and management and financial operations

Updated:

Models tested: 151

Human pairwise preferences on industry business and management and financial operations prompts in Arena’s text evaluation. The original organizer publishes model ratings and…

Top models · Arena

Latest result per model · Version not reported

  1. 1Gemini 4 ArgonHigh1,524.59
  2. 2Claude Opus 5Max1,500.65
  3. 3Claude Opus 4.6High1,497.22
View details →

Arena text / industry legal and government

Updated:

Models tested: 151

Human pairwise preferences on industry legal and government prompts in Arena’s text evaluation. The original organizer publishes model ratings and uncertainty; these are…

Top models · Arena

Latest result per model · Version not reported

  1. 1Gemini 4 ArgonHigh1,548.16
  2. 2Claude Opus 5Max1,529.67
  3. 3Claude Opus 5High1,522.37
View details →

Arena text / industry medicine and healthcare

Updated:

Models tested: 151

Human pairwise preferences on industry medicine and healthcare prompts in Arena’s text evaluation. The original organizer publishes model ratings and uncertainty; these are…

Top models · Arena

Latest result per model · Version not reported

  1. 1Gemini 4 ArgonHigh1,554.76
  2. 2Claude Fable 5.1Max1,531.47
  3. 3Claude Opus 5Max1,525.42
View details →

Arena text style control / industry business and management and financial operations

Updated:

Models tested: 151

Human pairwise preferences on industry business and management and financial operations prompts in Arena’s text style control evaluation. The original organizer publishes model…

Top models · Arena

Latest result per model · Version not reported

  1. 1Gemini 4 ArgonHigh1,520.77
  2. 2Claude Opus 4.7High1,507.73
  3. 3Muse Spark 1.2Extra high1,502.36
View details →

Arena text style control / industry legal and government

Updated:

Models tested: 151

Human pairwise preferences on industry legal and government prompts in Arena’s text style control evaluation. The original organizer publishes model ratings and uncertainty,…

Top models · Arena

Latest result per model · Version not reported

  1. 1Gemini 4 ArgonHigh1,535.54
  2. 2Muse Spark 1.2Extra high1,526.66
  3. 3Claude Fable 5High1,510.88
View details →

Arena text style control / industry medicine and healthcare

Updated:

Models tested: 151

Human pairwise preferences on industry medicine and healthcare prompts in Arena’s text style control evaluation. The original organizer publishes model ratings and uncertainty,…

Top models · Arena

Latest result per model · Version not reported

  1. 1Gemini 4 ArgonHigh1,551.3
  2. 2Muse WonderMax1,525.76
  3. 3Claude Fable 5.1Max1,524.68
View details →

Methodology and provenance

Benchize collects and structures public benchmark reports. Scores, costs, timings, ranks and uncertainty remain attributed to their original sources. Different versions, configurations and measurement units are separate results. Missing fields stay unavailable.