DeepSeek V3.2 Speciale

DeepSeek-V3.2-Speciale is a high-compute variant of DeepSeek-V3.2 optimized for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context processing, then scales post-training reinforcement learning...

by DeepSeek

Overview

Quick stats across all benchmark runs.

Score

52%

57 benchmarks

Avg Latency

149.1s

388 requests

Pricing

$0.29 in / $0.43 out

per 1M tokens

Context

164K

tokens

Alternatives

Models with similar or better quality but different tradeoffs

Same Quality, Cheaper

Models with similar or better performance at a lower cost per token.

Model	Cost
Gemini 2.0 Flash	-90%
DeepSeek V3.2	-87%
DeepSeek V3 0324	-82%
Mistral Large 3 2512	-69%
Gemini 3 Flash Preview	-61%

Same Quality, Faster

Models with similar or better performance but lower latency.

Model	Latency
Gemini 3 Flash Preview	-94%
Gemini 2.0 Flash	-94%
Gemini 2.5 Flash	-93%
Claude Haiku 4.5	-93%
Claude Sonnet 4.5	-90%

Same Cost, Better

Models at a similar price point with higher benchmark scores.

Model	Score
Gemini 3 Flash Preview	+9%
GPT-5 Mini	+1%

Benchmark Performance

How this model performs across different benchmarks

Benchmark	Score	Rank
Spatial Reasoning: Germany	95%	5 / 29
Character Frequency Bench	91%	9 / 29
Explain Like I'm 5	76%	17 / 47
Niederstetten Benchmark	74%	5 / 35
Categorization Bench	49%	28 / 54
Money Boy Cultural Literacy Test	38%	69 / 107
German Memelord Bench	36%	12 / 29

Price vs Performance

Compare cost efficiency across all models

Current model (baseline)

Other models (relative score)

Y-axis shows score difference from shared benchmarks. X-axis uses log scale.

Score Over Time

Performance trends across all benchmark runs

Benchmark Activity

Number of benchmark runs over time

Quickstart

View on OpenRouter

Get started with this model using OpenRouter

import { OpenRouter } from "@openrouter/sdk";

const openrouter = new OpenRouter({
  apiKey: "<OPENROUTER_API_KEY>"
});

const completion = await openrouter.chat.completions.create({
  model: "deepseek/deepseek-v3.2-speciale",
  messages: [
    {
      role: "user",
      content: "Hello!"
    }
  ]
});

console.log(completion.choices[0].message.content);

Get your API key at openrouter.ai/keys

Other Models from DeepSeek

Compare performance with other models from the same creator

Model	Latency	Cost/1M	Score
DeepSeek V4 Pro	12.3s	$0.65	59%
DeepSeek V4 Flash	5.0s	$0.17	55%
DeepSeek V3.2	36.7s	$0.32	40%
DeepSeek V3.1	52.2s	$0.50	—
DeepSeek R1 0528 Qwen3 8B	—	$0.07	—
DeepSeek Prover V2	—	$1.34	—
R1 Distill Qwen 14B	—	$0.15	—
R1 Distill Qwen 32B	53.5s	$0.29	—
DeepSeek V3.1 Terminus	20.0s	$0.61	—
R1	109.7s	$1.60	—
R1 0528	120.7s	$1.32	—
DeepSeek V3.2 Exp	79.4s	$0.34	—
DeepSeek V3	19.8s	$0.60	—
R1 Distill Llama 70B	33.8s	$0.75	—
DeepSeek V3 0324	15.2s	$0.49	—