CorpFin
As of Nov 22, 2026, CorpFin (v2) is maintained by Vals AI. Metric accuracy (%). Ceiling 100. Saturation unknown. Contamination risk low.
- Maintainer
- Vals AI
- Metric
- accuracy (%)
- Ceiling
- 100
- Saturation
- unknown
- Contamination
- low
- Cadence
- Declared cadence: as Vals updates (checked daily). Catalog snapshots: 1, latest Nov 13, 2026.
- Source
- Vals AI · this benchmark on Vals AI
- Grading
- AI judge (LLM-graded rubric) — Answers on long credit agreements are graded by an LLM judge against reference answers.
Source: Vals AI. A proprietary private test set run by Vals; scores are republished with permission. Rows are dated the day isitsota first saw them, and only the benchmark's standard harness counts.
What CorpFin measures
CorpFin is maintained by Vals AI and reports accuracy (%), where higher is better. The maximum is 100. It keeps a private held-out set. Grading: AI judge (LLM-graded rubric) (Answers on long credit agreements are graded by an LLM judge against reference answers.). Contamination risk is rated low. License: Source: Vals AI. Proprietary private set, run by Vals. Grading: AI judge (LLM-graded rubric). Answers on long credit agreements are graded by an LLM judge against reference answers.
99 models have a current score; the median is 61.4. #1 and #10 are 4.9 apart. The top score is 26.8 below the ceiling.
Where it counts
- Long context — secondary, weight 0.35
Current top scores
- Claude Opus 5 · 73.194 · Nov 13, 2026
- Claude Fable 5 · 71.834 · Nov 13, 2026
- Kimi K3 · 71.562 · Nov 13, 2026
- Muse Spark 1.1 · 71.29 · Nov 13, 2026
- Muse Spark 1.2 · 70.94 · Nov 13, 2026
Who held the top score
- Nov 13, 2026: Claude Opus 5 · 73.2
| Model | Score | Date | Kind |
|---|---|---|---|
| Claude Opus 5 | 73.194 | 2026-11-13 | snapshot |
| Claude Fable 5 | 71.834 | 2026-11-13 | snapshot |
| Kimi K3 | 71.562 | 2026-11-13 | snapshot |
| Muse Spark 1.1 | 71.29 | 2026-11-13 | snapshot |
| Muse Spark 1.2 | 70.94 | 2026-11-13 | snapshot |
| Inkling Small | 69.619 | 2026-11-13 | snapshot |
| Inkling | 68.57 | 2026-11-13 | snapshot |
| Grok 4.3 | 68.532 | 2026-11-13 | snapshot |
| GPT-5.5 | 68.415 | 2026-11-13 | snapshot |
| Kimi K2 5 Thinking | 68.259 | 2026-11-13 | snapshot |
| Minimax M3 | 68.104 | 2026-11-13 | snapshot |
| Qwen3 Max | 68.026 | 2026-11-13 | snapshot |
| Grok 4.5 | 67.405 | 2026-11-13 | snapshot |
| Claude Opus 4.6 Thinking | 67.016 | 2026-11-13 | snapshot |
| Claude Sonnet 5 | 66.977 | 2026-11-13 | snapshot |
| Grok 4 Fast Reasoning | 66.9 | 2026-11-13 | snapshot |
| Kimi K2 6 | 66.744 | 2026-11-13 | snapshot |
| Claude Opus 4.8 | 66.706 | 2026-11-13 | snapshot |
| Qwen3 6 Max Preview | 66.473 | 2026-11-13 | snapshot |
| Gemini 3 Flash Preview | 66.434 | 2026-11-13 | snapshot |
| Grok 4.6 | 66.161 | 2026-11-13 | snapshot |
| GLM 5.2 | 66.123 | 2026-11-13 | snapshot |
| Claude Opus 4.7 | 66.084 | 2026-11-13 | snapshot |
| Grok 4 0709 | 66.045 | 2026-11-13 | snapshot |
| Grok 4.1 Fast Reasoning | 65.967 | 2026-11-13 | snapshot |
| GPT-5.2 | 65.889 | 2026-11-13 | snapshot |
| Qwen3 8 Max | 65.851 | 2026-11-13 | snapshot |
| DeepSeek V4 Pro 0813 | 65.423 | 2026-11-13 | snapshot |
| GPT-5.6 Terra | 65.307 | 2026-11-13 | snapshot |
| Claude Sonnet 4.6 | 65.307 | 2026-11-13 | snapshot |