modelgrep
D

DeepSeek: DeepSeek V3.2

deepseek/deepseek-v3.2

38th smartest of 180ReasoningToolsJSON
Use via OpenRouter ↗
Intelligence
41.7
38th of 180
Design Elo
1221
Website
Speed
83
91st fastest
Latency
688ms
first token
Input price
$0.229
106th cheapest
Context
131K
64K max out

How it compares

Smarter than79%
of all ranked models
Faster than70%
of all ranked models
Cheaper than64%
of all ranked models

Overview

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...

Benchmarks

independent · via OpenRouter
Artificial Analysis76th percentile
Intelligence Index
41.7
Coding Index
36.7
Agentic Index
52.9
GPQA Diamond
84%
Humanity's Last Exam
22%
SciCode
39%
Tau²-Bench (agentic)
91%
Design Arena · Elo12,406 tournaments
Website
1221
codecategories
1217
3D
1214
UI Component
1213
Game Dev
1203
Data Viz
1203
asciiart
1109
svg
1087

Providers & pricing (11)

ProviderIn $/MOut $/MUptime
StreamLake$0.229$0.343100%
Baidufp8$0.252$0.378100%
SiliconFlowfp8$0.259$0.42099.6%
DeepInfrafp4$0.260$0.38099%
AtlasCloudfp8$0.260$0.38099.9%
Novitafp8$0.269$0.400100%
Alibaba$0.370$1.1198.8%
DigitalOcean$0.425$1.3699.8%
Friendli$0.500$1.5099.9%
Google$0.560$1.6899.2%
SambaNova$3.00$4.5098%

Specifications

Context window131K
Max output64K
Knowledge cutoff
Input modalitiestext
Output modalitiestext
Prompt caching
Cache read price
ModeratedNo

DeepSeek V3.2 FAQ

How much does DeepSeek V3.2 cost?

DeepSeek V3.2 costs $0.229 per million input tokens and $0.343 per million output tokens via OpenRouter, making it 106th cheapest of 298 paid models.

How smart is DeepSeek V3.2?

DeepSeek V3.2 scores 41.7 on the Artificial Analysis Intelligence Index, ranking 38th of 180 benchmarked models, with a GPQA Diamond score of 84%.

How fast is DeepSeek V3.2?

DeepSeek V3.2 generates around 83 tokens per second with 688ms time-to-first-token (p50), the 91st fastest tracked model.

What is DeepSeek V3.2's context window?

DeepSeek V3.2 supports a 131K-token context window and can output up to 64K tokens. It accepts text input.

Compare head-to-head