FinBalance: leaderboard
Metric: Exact balance-sheet accuracy (%): share of records whose reported final balance sheet matches the ledger ground truth exactly (BS exact); 480 standard records of the 710-record core evaluation split (source-document bundles with distractors across eight industries, three period types and five difficulty levels), baseline prompt, temperature 0, one pass; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 (Low) | 46.5 |
| 2 | Grok 4.3 (Non-reasoning) | 31 |
| 3 | Gemini 3 Flash (Minimal) | 23.5 |
| 4 | Claude Haiku 4.5 | 2.9 |
| 5 | Qwen 3 235B A22B 2507 Instruct | 1.3 |
Interactive version: theaggregate.ai/benchmark?slug=finbalance · How It Works · Data refreshed daily, snapshot 2026-09-29.