DeepSeek: DeepSeek V3.2 Exp

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-gra…

deepseek/deepseek-v3.2-exp

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve training and inference efficiency in long-context scenarios while maintaining output quality. Users can control the reasoning behaviour with the `reasoning` `enabled` boolean. [Learn more in our docs](https://openrouter.ai/docs/use-cases/reasoning-tokens#enable-reasoning-with-default-config)

The model was trained under conditions aligned with V3.1-Terminus to enable direct comparison. Benchmarking shows performance roughly on par with V3.1 across reasoning, coding, and agentic tool-use tasks, with minor tradeoffs and gains depending on the domain. This release focuses on validating architectural optimizations for extended context lengths rather than advancing raw task accuracy, making it primarily a research-oriented model for exploring efficient transformer designs.

Specifications & capabilities

Input
text
Output
text
Context window · tokens
163840
Knowledge cutoff
2025-07-31
Published model ID
DeepSeek-V3.2-Exp
Capabilities
reasoning, structuredOutputs, toolCalling
Description source ↗

Evaluations

81–100 / 102

LMArena

1,436.19

source id
lmarena
leaderboard name
LMArena Text Industry Software And It Services
dimension
general
metric
Arena rating
rank
132
sample size
4136
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1426.748521063746
confidence upper
1445.630612734193
Original result ↗

LMArena

1,435.328

source id
lmarena
leaderboard name
LMArena Text Coding
dimension
general
metric
Arena rating
rank
132
sample size
2541
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1423.776221520077
confidence upper
1446.880719763684
Original result ↗

LMArena

1,421.936

source id
lmarena
leaderboard name
LMArena Text Style Control Overall
dimension
style
metric
Arena rating
rank
133
sample size
11955
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1415.5489627734544
confidence upper
1428.3220994865308
Original result ↗

LMArena

1,406.069

source id
lmarena
leaderboard name
LMArena Text Style Control Non English
dimension
style
metric
Arena rating
rank
133
sample size
6901
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1398.2233581969842
confidence upper
1413.9140865681554
Original result ↗

LMArena

1,467.097

source id
lmarena
leaderboard name
LMArena Text Style Control Coding
dimension
style
metric
Arena rating
rank
134
sample size
2541
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1455.4836649420017
confidence upper
1478.710645856846
Original result ↗

LMArena

1,436.695

source id
lmarena
leaderboard name
LMArena Text Style Control Industry Life And Physical And Social Science
dimension
style
metric
Arena rating
rank
137
sample size
1930
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1423.2997475832635
confidence upper
1450.089442538924
Original result ↗

LMArena

1,454.056

source id
lmarena
leaderboard name
LMArena Text Style Control Industry Software And It Services
dimension
style
metric
Arena rating
rank
139
sample size
4136
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1444.5611833494124
confidence upper
1463.5515368359809
Original result ↗

LMArena

1,270.407

source id
lmarena
leaderboard name
LMArena Webdev Webdev
dimension
coding
metric
Arena rating
rank
139
sample size
3902
evaluated at
2026-11-10T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1256.7877189515946
confidence upper
1284.0263110593535
Original result ↗

LMArena

1,270.407

source id
lmarena
leaderboard name
LMArena Webdev Overall
dimension
coding
metric
Arena rating
rank
139
sample size
3902
evaluated at
2026-11-10T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1256.7877189515946
confidence upper
1284.0263110593535
Original result ↗

LMArena

1,434.387

source id
lmarena
leaderboard name
LMArena Text Style Control Industry Medicine And Healthcare
dimension
style
metric
Arena rating
rank
140
sample size
581
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1410.4433640225107
confidence upper
1458.3300682969718
Original result ↗

LMArena

1,279.541

source id
lmarena
leaderboard name
LMArena Webdev Webdev Html
dimension
coding
metric
Arena rating
rank
141
sample size
3902
evaluated at
2026-11-10T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1264.5478346855903
confidence upper
1294.5342997789292
Original result ↗

LMArena

1,344.38

source id
lmarena
leaderboard name
LMArena Text Style Control Japanese
dimension
style
metric
Arena rating
rank
143
sample size
148
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1296.1103331654795
confidence upper
1392.6497935946836
Original result ↗

LMArena

1,414.933

source id
lmarena
leaderboard name
LMArena Text Industry Mathematical
dimension
general
metric
Arena rating
rank
144
sample size
635
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1392.5988608885878
confidence upper
1437.2661546561594
Original result ↗

LMArena

1,331.505

source id
lmarena
leaderboard name
LMArena Text Japanese
dimension
general
metric
Arena rating
rank
147
sample size
148
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1282.0484346588369
confidence upper
1380.9625093897125
Original result ↗

LMArena

1,409.411

source id
lmarena
leaderboard name
LMArena Text Style Control Math
dimension
style
metric
Arena rating
rank
148
sample size
783
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1389.2344772127713
confidence upper
1429.5873412953688
Original result ↗

LMArena

1,403.724

source id
lmarena
leaderboard name
LMArena Text Style Control Russian
dimension
style
metric
Arena rating
rank
152
sample size
1170
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1387.077715105907
confidence upper
1420.3693575912152
Original result ↗

LMArena

1,406.804

source id
lmarena
leaderboard name
LMArena Text Industry Legal And Government
dimension
general
metric
Arena rating
rank
153
sample size
793
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1386.4497375810267
confidence upper
1427.1579547235287
Original result ↗

LMArena

1,414.554

source id
lmarena
leaderboard name
LMArena Text Style Control Industry Legal And Government
dimension
style
metric
Arena rating
rank
155
sample size
793
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1394.1300557367877
confidence upper
1434.978680294896
Original result ↗

LMArena

1,425.383

source id
lmarena
leaderboard name
LMArena Text Style Control Expert
dimension
style
metric
Arena rating
rank
157
sample size
661
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1403.1341613144982
confidence upper
1447.6323072395198
Original result ↗

LMArena

1,403.7

source id
lmarena
leaderboard name
LMArena Text Expert
dimension
general
metric
Arena rating
rank
157
sample size
661
evaluated at
2026-11-09T00:00:00.000Z
source model id
DeepSeek-V3.2-Exp
confidence lower
1381.543571221671
confidence upper
1425.8561653566762
Original result ↗

Published prices

1–6 / 6

deepseek-ai/DeepSeek-V3.2-Exp ↗

$0.27

Inference provider
siliconFlow
Platform
SiliconFlow
Unit
input_tokens_per_million
Currency
USD
Billing conditions
Billing label
Input tokens (per 1M)

deepseek-ai/DeepSeek-V3.2-Exp ↗

$0.41

Inference provider
siliconFlow
Platform
SiliconFlow
Unit
output_tokens_per_million
Currency
USD
Billing conditions
Billing label
Output tokens (per 1M)

Provider specifications & access

1–3 / 3

DeepSeek-V3.2-Exp ↗

source model id
deepseek-ai/DeepSeek-V3.2-Exp
hosting
siliconflow-hosted
platforms
  • SiliconFlow
max output tokens
163840
architecture
input_modalities
  • text
output_modalities
  • text
capabilities
  • reasoning
  • structuredOutputs
  • toolCalling

deepseek-ai/DeepSeek-V3.2-Exp ↗

source model id
publishers/deepseek-ai/models/deepseek-v3-2@deepseek-v3-2-exp
hosting
vertex-self-deployed
platforms
  • Google Cloud Vertex AI
architecture
input_modalities
  • text
output_modalities
  • text
capabilities

    deepseek-ai/DeepSeek-V3.2-Exp ↗

    source model id
    publishers/hf-deepseek-ai/models/deepseek-v3.2-exp@001
    hosting
    vertex-self-deployed
    platforms
    • Google Cloud Vertex AI
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities

      Safety assessments

      1–1 / 1

      Foundation Model Transparency Index 2025 ↗

      32 / 100

      source id
      stanfordFMTI
      canonical model id
      deepseek/deepseek-v3.2-exp
      taxonomy
      transparency
      hazard category
      transparency
      hazard name
      Organization transparency
      policy area
      DeepSeek
      assessment type
      organization-level disclosure
      metric
      disclosure_score
      result label
      32 / 100
      sample size
      100
      updated at
      2026-11-09T18:08:05.227Z

      Source specification observations

      1–5 / 5

      openRouter ↗

      field
      context_length
      value
      163840
      normalized value
      163840
      scope
      source-model
      source id
      openRouter
      confidence
      1

      siliconFlow ↗

      field
      context_length
      value
      164K
      scope
      source-model
      source id
      siliconFlow
      confidence
      1

      openRouter ↗

      field
      knowledge_cutoff
      reported field
      knowledge_cutoff
      value
      2025-07-31
      normalized value
      2025-07-31
      precision
      day
      scope
      source-model
      source id
      openRouter
      confidence
      1

      openRouter ↗

      field
      max_output_tokens
      value
      147456
      normalized value
      147456
      scope
      source-model
      source id
      openRouter
      confidence
      1

      siliconFlow ↗

      field
      max_output_tokens
      value
      164K
      scope
      source-model
      source id
      siliconFlow
      confidence
      1

      Original sources

      1–19 / 19

      DeepSeek · Model reference ↗

      SiliconFlow · Model page ↗

      Hugging Face · Model repository ↗

      DeepInfra · Model page ↗

      Google Cloud · Model page ↗

      Google Cloud · Model page ↗

      OpenRouter · Model page ↗

      SiliconFlow · Playground ↗

      SiliconFlow · API documentation ↗

      GitHub · Model reference ↗

      DeepInfra · Catalog ↗

      Epoch AI · Catalog ↗

      GitHub · Catalog ↗

      GitHub · Catalog ↗

      Google Cloud · Catalog ↗

      LMArena · Catalog ↗

      OpenRouter · Catalog ↗

      SiliconFlow · Catalog ↗

      gorilla.cs.berkeley.edu · Catalog ↗

      Source links

      1–11 / 11

      Explore more models

      Models

      Loading
      EVIDENCE / COMPARISON

      Benchmarks

      Explore every benchmark

      Loading

      These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

      Install TensorTrace