---
title: "DeepSeek V3.1 840B Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for DeepSeek V3.1."
source: "https://rankedagi.com/models/deepseek-v3-1"
---

# DeepSeek V3.1

| Field | Value |
| --- | --- |
| Organization | DeepSeek |
| License | Open Source |
| Type | Non-thinking |
| Size | 840B |
| Version | Latest |
| Released | 2025-08-21 |
| Context window | 128K |
| Input cost per million tokens | $0.56 |
| Output cost per million tokens | $1.68 |
| Cached input cost per million tokens | $0.07 |
| Model ID | deepseek-chat |
| Last updated | 2026-05-03T23:59:41.186Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 51.3% | RankedAGI Coding Score |  |
| RankedAGI Agentic | Agents | 42.6% | RankedAGI Agentic Score |  |
| LiveCodeBench v5 | coding | 56.4% | LiveCodeBench - Evaluation of Large Language Models for Code |  |
| Aider Polyglot | coding | 68.4% | Aider Polyglot Code Completion Benchmark |  |
| RankedAGI Reasoning | reasoning | 48.5% | RankedAGI Reasoning Score |  |
| Humanity's Last Exam | reasoning | 15.9% | Multidisciplinary Reasoning (no tools) |  |
| Text Arena | general | 1418 | ChatArena (LMSYS) ELO Score | https://arena.ai/leaderboard/text |
| AIME 2025 | math | 49.8% | AIME 2025 Competition Math |  |
| AIME 2024 | math | 66.3% | AIME 2024 Competition Math |  |
| MMLU Pro | knowledge | 83.7% | Graduate-level visual problem-solving |  |
| MMLU | knowledge | 91.8% | Image Reasoning |  |
| RankedAGI Math | math | 61.4% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 49.5% | Overall RankedAGI score |  |
| GDPval-AA Elo | Agents | 978 | Office Tasks (Artificial Analysis) |  |
| Code DesignArena | Design | 1176 |  | https://www.designarena.ai/leaderboard/code |

## Model Links

| Title | URL |
| --- | --- |
| Release | https://x.com/deepseek_ai/status/1958417062008918312 |
| DeepSeek | https://api-docs.deepseek.com/news/news250821 |