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

# DeepSeek V3.2 Speciale

| 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 | $0.4 |
| Output cost per million tokens | $1.2 |
| Cached input cost per million tokens | $0.2 |
| Model ID | deepseek-v3.2-speciale |
| Last updated | 2026-04-20T06:15:30.547Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 65.8% | RankedAGI Coding Score |  |
| RankedAGI Agentic | Agents | 59.6% | RankedAGI Agentic Score |  |
| SvelteBench | coding | 87.8% | SvelteBench - Benchmark for Svelte | https://khromov.github.io/svelte-bench/benchmark-results-merged.html |
| LiveCodeBench v6 | coding | 88.7% | LiveCodeBench v6 - Evaluation of Large Language Models for Code |  |
| Codeforces ELO Score | coding | 2701 | Codeforces ELO Score |  |
| RankedAGI Reasoning | reasoning | 67.0% | RankedAGI Reasoning Score |  |
| Humanity's Last Exam | reasoning | 30.6% | Multidisciplinary Reasoning (no tools) |  |
| GPQA Diamond | reasoning | 85.7% | Generalized Prefix Question Answering Score (Reasoning) PhD Level Reasoning |  |
| AIME 2025 | math | 96.0% | AIME 2025 Competition Math |  |
| RankedAGI Math | math | 75.7% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 60.6% | Overall RankedAGI score |  |
| GDPval-AA Elo | Agents | 500 | Office Tasks (Artificial Analysis) |  |

## Model Links

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