# DeepSeek

DeepSeek is a privately held technology company founded in 2016, headquartered in New York with a presence in Poland, that develops AI models in the DeepSeek-V3.x series featuring DeepSeek Sparse Attention for efficient long-context inference and achieving gold-level results in international competitive benchmarks.

Status: provisional
Last verified: 2026-08-29T06:09:24.000Z

## Evidence

- (single_source) DeepSeek is headquartered in New York, New York, United States, with a presence in Poland.
  - supports https://deepseek.com/
    > DeepSeek is a Technology, Information and Internet company. DeepSeek employs 87 people, founded in 2016. Headquartered in New York, New York, United States, with presence in Poland. Its workforce is distributed across 38 countries (including United States, Nigeria, India, Turkey, and Indonesia).
...
(AGI). We focus on building high-performance, efficient, and accessible AI models that empower
...
- Industry: Technology, Information and Internet
- Type: Privately Held
- Headquarters: New York, United States
- Founded Year: 2016
- Homepage: deepseek.com
- Status: active
- Aliases: deepseek, markets able, chaincreator
- LinkedIn: linkedin.com/company/deep-seek-ai-model
- Linkedin Followers: 18,616 (Monthly: +1.0%)
- Emails: contact@marketsable.com
- SIC Codes: 62, 623
- NAICS Codes: 52, 523

- (single_source) DeepSeek-V3.1-Terminus includes agent upgrades with stronger Code Agent and Search Agent performance.
  - supports https://deepseek.com/en/news/v3-1-terminus
    > DeepSeek | DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
# DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
The latest update builds on V3.1's strengths while addressing key user feedback.
...
- Language consistency: fewer CN/EN mix-ups & no more random chars.
- Agent upgrades: stronger Code Agent & Search Agent performance.
...
DeepSeek-V3.1-Terminus delivers more stable & reliable outputs across benchmarks compared to the previous version.
...
Available now on: App / Web / API
...
Open-source weights here: DeepSeek-V3.1-Terminus - Hugging Face
...
### DeepSeek-V3.1 Release
...
DeepSeek-V3.1 introduces hybrid Think & Non-Think inference, faster thinking, stronger agent skills, Anthropic API support, and 128K context.Read more
...
### DeepSeek-V3.2:
...
DeepSeek-V3.2
...
V3.2

- (single_source) DeepSeek is a privately held company with an active status.
  - supports https://deepseek.com/
    > DeepSeek is a Technology, Information and Internet company. DeepSeek employs 87 people, founded in 2016. Headquartered in New York, New York, United States, with presence in Poland. Its workforce is distributed across 38 countries (including United States, Nigeria, India, Turkey, and Indonesia).
...
(AGI). We focus on building high-performance, efficient, and accessible AI models that empower
...
- Industry: Technology, Information and Internet
- Type: Privately Held
- Headquarters: New York, United States
- Founded Year: 2016
- Homepage: deepseek.com
- Status: active
- Aliases: deepseek, markets able, chaincreator
- LinkedIn: linkedin.com/company/deep-seek-ai-model
- Linkedin Followers: 18,616 (Monthly: +1.0%)
- Emails: contact@marketsable.com
- SIC Codes: 62, 623
- NAICS Codes: 52, 523

- (single_source) DeepSeek-V3.2-Speciale attains gold-level results in IMO, CMO, ICPC World Finals, and IOI.
  - supports https://deepseek.com/en/news/deepseek-v3-2
    > # DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents!
...
- DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
- DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.
...
-3.0-Pro.
- Gold-Medal Performance: V3.2-Speciale attains gold-level results in IMO, CMO, ICPC World Finals & IOI
...
- V3.2: Same usage pattern as V3.2-Exp.
- V3.2-Speciale: Served via a temporary endpoint:`base_url="https://api.deepseek.com/v3.2_speciale_expires_on_20251215"`. Same pricing as V3.2, no tool calls, available until Dec 15th, 2025, 15:59 (UTC Time).
- V3.2 now supports Thinking in Tool-Use — details: Thinking Mode Guide - API Docs

- (single_source) DeepSeek-V3.2 now supports Thinking in Tool-Use.
  - supports https://deepseek.com/en/news/deepseek-v3-2
    > # DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents!
...
- DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
- DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.
...
-3.0-Pro.
- Gold-Medal Performance: V3.2-Speciale attains gold-level results in IMO, CMO, ICPC World Finals & IOI
...
- V3.2: Same usage pattern as V3.2-Exp.
- V3.2-Speciale: Served via a temporary endpoint:`base_url="https://api.deepseek.com/v3.2_speciale_expires_on_20251215"`. Same pricing as V3.2, no tool calls, available until Dec 15th, 2025, 15:59 (UTC Time).
- V3.2 now supports Thinking in Tool-Use — details: Thinking Mode Guide - API Docs

- (single_source) Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.
  - supports https://deepseek.com/en/news/v3-2-exp
    > # Introducing DeepSeek-V3.2-Exp
...
Introducing DeepSeek-V3.2-Exp — our latest experimental model!
...
Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention (DSA) for faster, more efficient training & inference on long context.
...
- DSA achieves fine-grained sparse attention with minimal impact on output quality — boosting long-context performance & reducing compute cost.
...
- Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.
...
- DeepSeek-V3.2-Exp - Hug
...
- DeepSeek-V3.2
...
### DeepSeek
...
V3.1 is now DeepSeek-V3.1-Terminus
...
### DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
DeepSeek-V3.2 and V3.2-Speciale launch with world-leading reasoning, thinking in tool-use, and gold-medal performance in IMO, CMO, ICPC & IOI 2025.Read more

- (single_source) DeepSeek-V3.1 is now named DeepSeek-V3.1-Terminus.
  - supports https://deepseek.com/en/news/v3-2-exp
    > # Introducing DeepSeek-V3.2-Exp
...
Introducing DeepSeek-V3.2-Exp — our latest experimental model!
...
Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention (DSA) for faster, more efficient training & inference on long context.
...
- DSA achieves fine-grained sparse attention with minimal impact on output quality — boosting long-context performance & reducing compute cost.
...
- Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.
...
- DeepSeek-V3.2-Exp - Hug
...
- DeepSeek-V3.2
...
### DeepSeek
...
V3.1 is now DeepSeek-V3.1-Terminus
...
### DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
DeepSeek-V3.2 and V3.2-Speciale launch with world-leading reasoning, thinking in tool-use, and gold-medal performance in IMO, CMO, ICPC & IOI 2025.Read more

- (single_source) DeepSeek-V4 uses novel attention combining token-wise compression and DSA (DeepSeek Sparse Attention).
  - supports https://deepseek.com/en/news/v4-preview
    > DeepSeek-V
...
DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
...
- DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
- DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
...
## DeepSeek-V4-Pro
...
## DeepSeek-V4-Flash
...
- Novel Attention: Token-wise compression + DSA (DeepSeek Sparse Attention).
- Peak Efficiency: World-leading long context with drastically reduced compute & memory costs.
- 1M Standard: 1M context is now the default across all official DeepSeek services.
...
## API is Available Today!
...
- Keep base_url, just update model to`deepseek-v4-pro` or`deepseek-v4-flash`.
- Supports OpenAI ChatComple

- (single_source) DeepSeek is a Technology, Information and Internet company that employs 87 people and was founded in 2016.
  - supports https://deepseek.com/
    > DeepSeek is a Technology, Information and Internet company. DeepSeek employs 87 people, founded in 2016. Headquartered in New York, New York, United States, with presence in Poland. Its workforce is distributed across 38 countries (including United States, Nigeria, India, Turkey, and Indonesia).
...
(AGI). We focus on building high-performance, efficient, and accessible AI models that empower
...
- Industry: Technology, Information and Internet
- Type: Privately Held
- Headquarters: New York, United States
- Founded Year: 2016
- Homepage: deepseek.com
- Status: active
- Aliases: deepseek, markets able, chaincreator
- LinkedIn: linkedin.com/company/deep-seek-ai-model
- Linkedin Followers: 18,616 (Monthly: +1.0%)
- Emails: contact@marketsable.com
- SIC Codes: 62, 623
- NAICS Codes: 52, 523

- (single_source) DeepSeek-V4 Preview is officially live and open-sourced, introducing cost-effective 1M context length.
  - supports https://deepseek.com/en/news/v4-preview
    > DeepSeek-V
...
DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
...
- DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
- DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
...
## DeepSeek-V4-Pro
...
## DeepSeek-V4-Flash
...
- Novel Attention: Token-wise compression + DSA (DeepSeek Sparse Attention).
- Peak Efficiency: World-leading long context with drastically reduced compute & memory costs.
- 1M Standard: 1M context is now the default across all official DeepSeek services.
...
## API is Available Today!
...
- Keep base_url, just update model to`deepseek-v4-pro` or`deepseek-v4-flash`.
- Supports OpenAI ChatComple

- (single_source) DeepSeek-V4-Flash has 284B total parameters and 13B active parameters, positioned as a fast, efficient, and economical choice.
  - supports https://deepseek.com/en/news/v4-preview
    > DeepSeek-V
...
DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
...
- DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
- DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
...
## DeepSeek-V4-Pro
...
## DeepSeek-V4-Flash
...
- Novel Attention: Token-wise compression + DSA (DeepSeek Sparse Attention).
- Peak Efficiency: World-leading long context with drastically reduced compute & memory costs.
- 1M Standard: 1M context is now the default across all official DeepSeek services.
...
## API is Available Today!
...
- Keep base_url, just update model to`deepseek-v4-pro` or`deepseek-v4-flash`.
- Supports OpenAI ChatComple

- (single_source) 1M context length is now the default across all official DeepSeek services for DeepSeek-V4.
  - supports https://deepseek.com/en/news/v4-preview
    > DeepSeek-V
...
DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
...
- DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
- DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
...
## DeepSeek-V4-Pro
...
## DeepSeek-V4-Flash
...
- Novel Attention: Token-wise compression + DSA (DeepSeek Sparse Attention).
- Peak Efficiency: World-leading long context with drastically reduced compute & memory costs.
- 1M Standard: 1M context is now the default across all official DeepSeek services.
...
## API is Available Today!
...
- Keep base_url, just update model to`deepseek-v4-pro` or`deepseek-v4-flash`.
- Supports OpenAI ChatComple

- (single_source) DSA achieves fine-grained sparse attention with minimal impact on output quality, boosting long-context performance and reducing compute cost.
  - supports https://deepseek.com/en/news/v3-2-exp
    > # Introducing DeepSeek-V3.2-Exp
...
Introducing DeepSeek-V3.2-Exp — our latest experimental model!
...
Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention (DSA) for faster, more efficient training & inference on long context.
...
- DSA achieves fine-grained sparse attention with minimal impact on output quality — boosting long-context performance & reducing compute cost.
...
- Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.
...
- DeepSeek-V3.2-Exp - Hug
...
- DeepSeek-V3.2
...
### DeepSeek
...
V3.1 is now DeepSeek-V3.1-Terminus
...
### DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
DeepSeek-V3.2 and V3.2-Speciale launch with world-leading reasoning, thinking in tool-use, and gold-medal performance in IMO, CMO, ICPC & IOI 2025.Read more

- (single_source) DeepSeek-V3.2 is the official successor to V3.2-Exp and is available on App, Web, and API.
  - supports https://deepseek.com/en/news/deepseek-v3-2
    > # DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents!
...
- DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
- DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.
...
-3.0-Pro.
- Gold-Medal Performance: V3.2-Speciale attains gold-level results in IMO, CMO, ICPC World Finals & IOI
...
- V3.2: Same usage pattern as V3.2-Exp.
- V3.2-Speciale: Served via a temporary endpoint:`base_url="https://api.deepseek.com/v3.2_speciale_expires_on_20251215"`. Same pricing as V3.2, no tool calls, available until Dec 15th, 2025, 15:59 (UTC Time).
- V3.2 now supports Thinking in Tool-Use — details: Thinking Mode Guide - API Docs

- (single_source) DeepSeek-V3.1-Terminus is available on App, Web, and API, with open-source weights on Hugging Face.
  - supports https://deepseek.com/en/news/v3-1-terminus
    > DeepSeek | DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
# DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
The latest update builds on V3.1's strengths while addressing key user feedback.
...
- Language consistency: fewer CN/EN mix-ups & no more random chars.
- Agent upgrades: stronger Code Agent & Search Agent performance.
...
DeepSeek-V3.1-Terminus delivers more stable & reliable outputs across benchmarks compared to the previous version.
...
Available now on: App / Web / API
...
Open-source weights here: DeepSeek-V3.1-Terminus - Hugging Face
...
### DeepSeek-V3.1 Release
...
DeepSeek-V3.1 introduces hybrid Think & Non-Think inference, faster thinking, stronger agent skills, Anthropic API support, and 128K context.Read more
...
### DeepSeek-V3.2:
...
DeepSeek-V3.2
...
V3.2

- (single_source) DeepSeek-V4-Pro has 1.6T total parameters and 49B active parameters, with performance rivaling the world's top closed-source models.
  - supports https://deepseek.com/en/news/v4-preview
    > DeepSeek-V
...
DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
...
- DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
- DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
...
## DeepSeek-V4-Pro
...
## DeepSeek-V4-Flash
...
- Novel Attention: Token-wise compression + DSA (DeepSeek Sparse Attention).
- Peak Efficiency: World-leading long context with drastically reduced compute & memory costs.
- 1M Standard: 1M context is now the default across all official DeepSeek services.
...
## API is Available Today!
...
- Keep base_url, just update model to`deepseek-v4-pro` or`deepseek-v4-flash`.
- Supports OpenAI ChatComple

- (single_source) DeepSeek-V3.1-Terminus builds on V3.1's strengths while addressing key user feedback, including improved language consistency with fewer CN/EN mix-ups and no more random characters.
  - supports https://deepseek.com/en/news/v3-1-terminus
    > DeepSeek | DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
# DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
The latest update builds on V3.1's strengths while addressing key user feedback.
...
- Language consistency: fewer CN/EN mix-ups & no more random chars.
- Agent upgrades: stronger Code Agent & Search Agent performance.
...
DeepSeek-V3.1-Terminus delivers more stable & reliable outputs across benchmarks compared to the previous version.
...
Available now on: App / Web / API
...
Open-source weights here: DeepSeek-V3.1-Terminus - Hugging Face
...
### DeepSeek-V3.1 Release
...
DeepSeek-V3.1 introduces hybrid Think & Non-Think inference, faster thinking, stronger agent skills, Anthropic API support, and 128K context.Read more
...
### DeepSeek-V3.2:
...
DeepSeek-V3.2
...
V3.2

- (single_source) DeepSeek-V3.2 and V3.2-Speciale launch with world-leading reasoning, thinking in tool-use, and gold-medal performance in IMO, CMO, ICPC, and IOI 2025.
  - supports https://deepseek.com/en/news/v3-2-exp
    > # Introducing DeepSeek-V3.2-Exp
...
Introducing DeepSeek-V3.2-Exp — our latest experimental model!
...
Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention (DSA) for faster, more efficient training & inference on long context.
...
- DSA achieves fine-grained sparse attention with minimal impact on output quality — boosting long-context performance & reducing compute cost.
...
- Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.
...
- DeepSeek-V3.2-Exp - Hug
...
- DeepSeek-V3.2
...
### DeepSeek
...
V3.1 is now DeepSeek-V3.1-Terminus
...
### DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
DeepSeek-V3.2 and V3.2-Speciale launch with world-leading reasoning, thinking in tool-use, and gold-medal performance in IMO, CMO, ICPC & IOI 2025.Read more

- (single_source) DeepSeek-V3.2-Speciale is available via API only and is served through a temporary endpoint that expires on December 15, 2025 at 15:59 UTC.
  - supports https://deepseek.com/en/news/deepseek-v3-2
    > # DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents!
...
- DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
- DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.
...
-3.0-Pro.
- Gold-Medal Performance: V3.2-Speciale attains gold-level results in IMO, CMO, ICPC World Finals & IOI
...
- V3.2: Same usage pattern as V3.2-Exp.
- V3.2-Speciale: Served via a temporary endpoint:`base_url="https://api.deepseek.com/v3.2_speciale_expires_on_20251215"`. Same pricing as V3.2, no tool calls, available until Dec 15th, 2025, 15:59 (UTC Time).
- V3.2 now supports Thinking in Tool-Use — details: Thinking Mode Guide - API Docs

- (single_source) DeepSeek-V3.2-Exp is built on V3.1-Terminus and debuts DeepSeek Sparse Attention (DSA) for faster and more efficient training and inference on long context.
  - supports https://deepseek.com/en/news/v3-2-exp
    > # Introducing DeepSeek-V3.2-Exp
...
Introducing DeepSeek-V3.2-Exp — our latest experimental model!
...
Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention (DSA) for faster, more efficient training & inference on long context.
...
- DSA achieves fine-grained sparse attention with minimal impact on output quality — boosting long-context performance & reducing compute cost.
...
- Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.
...
- DeepSeek-V3.2-Exp - Hug
...
- DeepSeek-V3.2
...
### DeepSeek
...
V3.1 is now DeepSeek-V3.1-Terminus
...
### DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
...
DeepSeek-V3.2 and V3.2-Speciale launch with world-leading reasoning, thinking in tool-use, and gold-medal performance in IMO, CMO, ICPC & IOI 2025.Read more

- (single_source) DeepSeek-V4 API is available today, supporting model names deepseek-v4-pro and deepseek-v4-flash, and supports OpenAI ChatCompletion format.
  - supports https://deepseek.com/en/news/v4-preview
    > DeepSeek-V
...
DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
...
- DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
- DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
...
## DeepSeek-V4-Pro
...
## DeepSeek-V4-Flash
...
- Novel Attention: Token-wise compression + DSA (DeepSeek Sparse Attention).
- Peak Efficiency: World-leading long context with drastically reduced compute & memory costs.
- 1M Standard: 1M context is now the default across all official DeepSeek services.
...
## API is Available Today!
...
- Keep base_url, just update model to`deepseek-v4-pro` or`deepseek-v4-flash`.
- Supports OpenAI ChatComple

- (single_source) DeepSeek-V3.1 introduces hybrid Think and Non-Think inference, faster thinking, stronger agent skills, Anthropic API support, and 128K context.
  - supports https://deepseek.com/en/news/v3-1-terminus
    > DeepSeek | DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
# DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus
...
The latest update builds on V3.1's strengths while addressing key user feedback.
...
- Language consistency: fewer CN/EN mix-ups & no more random chars.
- Agent upgrades: stronger Code Agent & Search Agent performance.
...
DeepSeek-V3.1-Terminus delivers more stable & reliable outputs across benchmarks compared to the previous version.
...
Available now on: App / Web / API
...
Open-source weights here: DeepSeek-V3.1-Terminus - Hugging Face
...
### DeepSeek-V3.1 Release
...
DeepSeek-V3.1 introduces hybrid Think & Non-Think inference, faster thinking, stronger agent skills, Anthropic API support, and 128K context.Read more
...
### DeepSeek-V3.2:
...
DeepSeek-V3.2
...
V3.2

- (single_source) DeepSeek has 18,616 LinkedIn followers with a monthly growth rate of 1.0%.
  - supports https://deepseek.com/
    > DeepSeek is a Technology, Information and Internet company. DeepSeek employs 87 people, founded in 2016. Headquartered in New York, New York, United States, with presence in Poland. Its workforce is distributed across 38 countries (including United States, Nigeria, India, Turkey, and Indonesia).
...
(AGI). We focus on building high-performance, efficient, and accessible AI models that empower
...
- Industry: Technology, Information and Internet
- Type: Privately Held
- Headquarters: New York, United States
- Founded Year: 2016
- Homepage: deepseek.com
- Status: active
- Aliases: deepseek, markets able, chaincreator
- LinkedIn: linkedin.com/company/deep-seek-ai-model
- Linkedin Followers: 18,616 (Monthly: +1.0%)
- Emails: contact@marketsable.com
- SIC Codes: 62, 623
- NAICS Codes: 52, 523

## Timeline

- 2026-08-29T06:01:32.000Z: DeepSeek is a technology company employing 87 people, founded in 2016. DeepSeek is headquartered in New York, NY, with a presence in Poland. DeepSeek is a privately held company with active status. DeepSeek has 18,616 LinkedIn followers growing at 1.0% monthly. DeepSeek-V3.2 is the official successor to V3.2-Exp, available on App, Web, and API. DeepSeek-V3.2-Speciale is API-only via a temporary endpoint expiring December 15, 2025. DeepSeek-V3.2-Speciale achieves gold-level results in IMO, CMO, ICPC World Finals, and IOI. DeepSeek-V3.2 now supports Thinking in Tool-Use. DeepSeek-V3.2-Exp is built on V3.1-Terminus and introduces DeepSeek Sparse Attention (DSA) for efficient long-context training and inference. DSA provides fine-grained sparse attention with minimal quality impact, improving long-context performance and reducing compute cost. Benchmarks show V3.2-Exp performs on par with V3.1-Terminus. DeepSeek-V3.1 has been renamed DeepSeek-V3.1-Terminus.