---
title: "DeepSeek V3.2 685B Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for DeepSeek V3.2."
source: "https://rankedagi.com/models/deepseek-v3-2"
---

# DeepSeek V3.2

| Field | Value |
| --- | --- |
| Organization | DeepSeek |
| License | Open Source |
| Type | Thinking |
| Size | 685B |
| Version | Latest |
| Released | 2025-12-01 |
| Context window | 164K |
| Input cost per million tokens | $0.26 |
| Output cost per million tokens | $0.38 |
| Cached input cost per million tokens | $0.13 |
| Model ID | deepseek-v3.2 |
| Last updated | 2026-05-03T23:59:41.186Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 67.1% | RankedAGI Coding Score |  |
| SWEBench Multilingual | coding | 58.0% |  |  |
| SWEBench Verified | coding | 73.1% | Agentic Coding | https://www.swebench.com/index.html |
| Terminal-Bench 2.0 | coding | 46.4% | Agentic Terminal Coding | https://www.tbench.ai/leaderboard/terminal-bench/2.0 |
| RankedAGI Agentic | Agents | 54.7% | RankedAGI Agentic Score |  |
| Code Arena | coding | 1368 | Code Arena Elo Score | https://arena.ai/leaderboard/code |
| Livebench Coding 26.1 | coding | 75.7% | Livebench Coding 26.01.08 | https://livebench.ai/#/?Coding=a&sort=Coding+Average&variants=true&highunseenbias=true |
| SvelteBench | coding | 91.1% | SvelteBench - Benchmark for Svelte | https://khromov.github.io/svelte-bench/benchmark-results-merged.html |
| LiveCodeBench v6 | coding | 83.3% | LiveCodeBench v6 - Evaluation of Large Language Models for Code |  |
| Codeforces ELO Score | coding | 2386 | Codeforces ELO Score |  |
| RankedAGI Reasoning | reasoning | 60.3% | RankedAGI Reasoning Score |  |
| Humanity's Last Exam | reasoning | 25.1% | Multidisciplinary Reasoning (no tools) |  |
| GPQA Diamond | reasoning | 82.4% | Generalized Prefix Question Answering Score (Reasoning) PhD Level Reasoning |  |
| Text Arena | general | 1424 | ChatArena (LMSYS) ELO Score | https://arena.ai/leaderboard/text |
| AIME 2026 | math | 94.2% | AIME 2026 Competition Math | https://matharena.ai/?comp=aime--aime_2026&view=problem |
| AIME 2025 | math | 93.1% | AIME 2025 Competition Math |  |
| Vending Bench 2 | Agents | $1,034 | Benchmark for measuring AI model performance on running a business over long time horizons. Models are tasked with running a simulated vending machine business over a year and scored on their bank account balance at the end. | https://andonlabs.com/evals/vending-bench-2 |
| NYT Connections Extended | reasoning | 50.2% | NYT Connections Extended Version | https://github.com/lechmazur/nyt-connections/ |
| Hughes Hallucination Evaluation | safety | 6.3% | Computed using Vectara's Hallucination Evaluation Model, also known as HHEM. This evaluates how often an LLM introduces hallucinations when summarizing a document. | https://github.com/vectara/hallucination-leaderboard |
| RankedAGI Math | math | 75.4% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 58.7% | Overall RankedAGI score |  |
| GDPval-AA Elo | Agents | 1202 | Office Tasks (Artificial Analysis) |  |
| Code DesignArena | Design | 1235 |  | https://www.designarena.ai/leaderboard/code |
| Toolathlon | Agents | 35.2% | Benchmark to assess language agents' general tool use in realistic environments. | https://toolathlon.xyz/introduction |

## Model Links

| Title | URL |
| --- | --- |
| Release | https://api-docs.deepseek.com/news/news251201 |
| OpenRouter | https://openrouter.ai/deepseek/deepseek-v3.2 |