Vals Index
Updated 2/12/2026
Vals Index
Benchmark consisting of a weighted performance across finance, law and coding tasks. Showing the potential impact that LLM's can have on the economy.
Top Model:
Claude Opus 4.6 (Thinking)
Number of models tested
28
Updated 2/10/2026
Vals Multimodal Index
Benchmark consisting of a weighted performance across finance, law, coding, and education tasks. Showing the potential impact that LLM's can have on the economy.
Top Model:
Claude Opus 4.6 (Thinking)
Number of models tested
17
Legal Benchmarks
Updated 2/12/2026
CaseLaw (v2)
Private question-answer benchmark over Canadian court-cases.
Top Model:
GPT 5 Mini
Number of models tested
63
Updated 2/12/2026
LegalBench
Evaluating language models on a wide range of open source legal reasoning tasks.
Top Model:
Gemini 3 Pro (11/25)
Number of models tested
104
Finance Benchmarks
Updated 2/12/2026
CorpFin (v2)
A private benchmark evaluating understanding of long-context credit agreements
Top Model:
Kimi K2.5
Number of models tested
82
Updated 2/12/2026
Finance Agent v1.1
Evaluating agents on core financial analyst tasks
Top Model:
Claude Opus 4.6 (Thinking)
Number of models tested
30
Updated 2/3/2026
MortgageTax
Evaluating reading and understanding tax certificates as images
Top Model:
Gemini 3 Pro (11/25)
Number of models tested
58
Updated 2/12/2026
TaxEval (v2)
A Vals-created set of questions and responses to tax questions
Top Model:
Claude Opus 4.6 (Thinking)
Number of models tested
91
Healthcare Benchmarks
Updated 2/12/2026
MedQA
Evaluating language model bias in medical questions.
Top Model:
o1
Number of models tested
90
Math Benchmarks
Updated 2/12/2026
AIME
Challenging national math exam given to top high-school students
Top Model:
GPT 5.2
Number of models tested
82
Updated 2/13/2026
ProofBench
Can models write math proofs that are formally verified?
Top Model:
Claude Opus 4.6 (Thinking)
Number of models tested
15
Academic Benchmarks
Updated 2/12/2026
GPQA
Graduate-level Google-Proof Q&A benchmark evaluating models on questions that require deep reasoning.
Top Model:
Gemini 3 Pro (11/25)
Number of models tested
85
Updated 2/12/2026
MMLU Pro
Academic multiple-choice benchmark covering 14 subjects including STEM, humanities, and social sciences.
Top Model:
Gemini 3 Pro (11/25)
Number of models tested
84
Updated 2/3/2026
MMMU
Multimodal Multi-task Benchmark
Top Model:
Gemini 3 Flash (12/25)
Number of models tested
55
Education Benchmarks
Updated 2/3/2026
SAGE
Student Assessment with Generative Evaluation
Top Model:
Claude Opus 4.5 (Thinking)
Number of models tested
37
Coding Benchmarks
Updated 2/12/2026
IOI
International Olympiad in Informatics
Top Model:
GPT 5.2
Number of models tested
43
Updated 2/12/2026
LiveCodeBench
Our Implementation of the LiveCodeBench benchmark
Top Model:
GPT 5 Mini
Number of models tested
90
Updated 2/13/2026
SWE-bench
Solving production software engineering tasks
Top Model:
Claude Opus 4.6 (Thinking)
Number of models tested
50
Updated
NewTerminal-Bench 2.0
State-of-the-art set of difficult terminal-based tasks
Top Model:
Claude Opus 4.5 (Nonthinking)
Number of models tested
35
Updated 2/5/2026
NewVibe Code Bench
Can models build web applications from scratch?
Top Model:
GPT 5.2
Number of models tested
16
Beta Benchmarks
Updated 12/23/2025
NewPoker Agent
Which model can make the most money playing poker?
Top Model:
GPT 5.2
Number of models tested
17