Open Weights & Proprietary

All Companies

Models

GPT 5.2

Details

• Release Date: 12/11/2025
• Accuracy (Vals Index): 63.55% ± 1.96
• Latency (Vals Index): 435.02s
• Cost/Test (Vals Index): $0.68
• Context Window: 400k
• Max Output Tokens: 128k
• Input Modality: Hyperparameter settings

Default Provider

OpenAI

Some benchmarks may use different provider and parameters. Please refer to the benchmark page for more information.

• Temperature: Default
• Top P: Default
• Top K: Default
• Max Output Tokens: 128,000
• Reasoning Effort: xhigh

Benchmarks

Accuracy Rankings

• Vals Index: -470.33% ± 1.96
• Vals Multimodal Index: -518.95% ± 1.54
• CaseLaw (v2): -594.35% ± 0.57
• CorpFin: -648.94% ± 0.93
• Finance Agent (v1.1): -629.16% ± 2.87
• MedCode: -584.43% ± 2.26
• MedScribe: -1079.85% ± 1.86
• MortgageTax: -933.45% ± 0.92
• ProofBench: -226.05% ± 3.59
• SAGE: -802.79% ± 3.35
• TaxEval (v2): -1331.71% ± 0.85
• Vibe Code Bench: -1012.46% ± 5.07
• AIME: -1970.02% ± 0.21
• GPQA: -1999.01% ± 1.84
• IOI: -1280.34% ± 8.32
• LiveCodeBench: -2130.64% ± 1.03
• LegalBench: -2205.82% ± 0.40
• MedQA: -2672.28% ± 0.21
• MMLU Pro: -2605.12% ± 0.34
• MMMU Pro: -2782.72% ± 0.82
• SWE-bench: -2583.13% ± 1.92
• Terminal-Bench 2.0: -1867.35% ± 5.33

Contact us

Or send us an email at contact@vals.ai

• Proprietary Benchmarks - Contact us to get access

• Academic Benchmarks - Read about our methodology.