---
title: "DeepSeek V4 Pro (DeepSeek) - Financial AI Performance - FinanceBenchmark"
description: "Independent evaluation and benchmark scores for DeepSeek V4 Pro on financial reasoning."
canonical: "https://financebenchmark.ai/models/deepseek-v4-pro"
type: "text/markdown"
---
# DeepSeek V4 Pro

> Developed by **DeepSeek** · Released 2026-01 · License: open-weight

## Table of Contents
- [1. Model Profile & Summary](#profile)
- [2. FB Composite Score](#composite-score)
- [3. All Evaluated Benchmarks](#benchmarks)
- [4. Certification & Verification](#verification)

## 1. Model Profile & Summary {#profile}

- **Organization:** DeepSeek
- **Release Date:** 2026-01
- **License:** open-weight (Open-Weight)
- **Context Window:** 128K
- **FB Certified:** Standard Attribution

## 2. FB Composite Score {#composite-score}

**Composite FB Score: 70.1 / 100**

- Benchmarks Evaluated: **9 of 42**

## 3. All Evaluated Benchmarks {#benchmarks}

| Benchmark | Category | Overall Score | Eval Date | Source |
|:----------|:---------|:-------------:|:---------:|:------:|
| [Pave Interchange Fee Benchmark](https://financebenchmark.ai/benchmarks/pave.md) | verification, forensic | **65.0%** | 2026-05 | [Paper](https://github.com/Finance-Benchmark/pave-benchmark) |
| [FinanceBench](https://financebenchmark.ai/benchmarks/financebench.md) | documentQA | **85.0%** | 2026-03 | [Paper](https://arxiv.org/abs/2311.11944) |
| [FinBen](https://financebenchmark.ai/benchmarks/finben.md) | multiDomain | **78.0%** | 2026-02 | [Paper](https://arxiv.org/abs/2402.12659) |
| [FinQA](https://financebenchmark.ai/benchmarks/finqa.md) | numerical | **87.0%** | 2026-02 | [Paper](https://arxiv.org/abs/2109.00122) |
| [ConvFinQA](https://financebenchmark.ai/benchmarks/convfinqa.md) | documentQA | **83.0%** | 2026-03 | [Paper](https://arxiv.org/abs/2210.03849) |
| [Finance Agent Benchmark](https://financebenchmark.ai/benchmarks/finance-agent.md) | agentTasks | **44.1%** | 2026-09 | [Paper](https://vals.ai/benchmarks/fabv2) |
| [MultiHiertt](https://financebenchmark.ai/benchmarks/multihiertt.md) | numerical | **82.0%** | 2026-02 | [Paper](https://arxiv.org/abs/2206.01347) |
| [FinAuditing](https://financebenchmark.ai/benchmarks/finauditing.md) | forensic, compliance | **73.0%** | 2026-03 | [Paper](https://arxiv.org/abs/2510.08886) |
| [Excel Modeling Benchmark](https://financebenchmark.ai/benchmarks/emb.md) | documentQA, forensic | **51.6%** | 2026-09 | [Paper](https://vals.ai/benchmarks/emb) |

## 4. Certification & Verification {#verification}

Embed live verification badge:

```html
<img src="https://financebenchmark.ai/embed/badge/deepseek-v4-pro" alt="DeepSeek V4 Pro FinanceBenchmark Score" />
```