# DeepSeek-V3.2 finance AI benchmark results

Every finance benchmark score for DeepSeek-V3.2 by DeepSeek: 3 benchmarks, each with its rank, metric and source date. Scores are never combined across benchmarks.

## Board standings

DeepSeek-V3.2 is first on 0 benchmarks, top 3 on 0 benchmarks, scored on 3 benchmarks. Places on each benchmark's main board with 2 or more models; scores are never combined. https://benchmarks.finance/method/#standings

## Scores

| Benchmark | Rank | Score | Metric | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [Market-Bench](https://benchmarks.finance/benchmarks/afterquery-market-bench/) | 5 of 10 | 4575.6 | mean MAE | marketbench.ai leaderboard | date not stated |
| [PolyBench](https://benchmarks.finance/benchmarks/polybench/) | 4 of 7 | -11.4 | Confidence-Weighted Return (CWR) % | Paper Table 5 | 3 Apr 2026 |
| [FinSafetyBench](https://benchmarks.finance/benchmarks/finsafetybench/) | 2 of 2 | 89.5 | attack success rate % (financial crimes) | Paper Table 3 | 1 May 2026 |

Source: https://benchmarks.finance/models/deepseek-v3-2/

Created and maintained by Moneda.