DeepSeek
Open-weight research lab known for DeepSeek V4, V3, and R1 reasoning models.
DeepSeek releases competitive open-weight models—including DeepSeek V4 (Pro/Flash), DeepSeek V3, and the reasoning-focused DeepSeek R1—plus APIs (including experimental deepseek-v4-flash-vision-exp), chat, and DeepSeek Harness (MIT agent runtime, developer preview). It forced the industry to revisit cost/performance assumptions for frontier-class open models, especially for reasoning and coding workloads.
Why DeepSeek matters
DeepSeek is the open-weight shock of the mid-2020s: R1-level reasoning and V4 agentic coding at a fraction of closed-API prices. If you are evaluating self-host, distillation, or whether to stay on GPT/Claude, DeepSeek is the comparison that often changes the spreadsheet.
Last reviewed: 24 August 2026
Best fit for
When architects typically choose DeepSeek.
- Open-weight reasoning models
- Cost-sensitive frontier alternatives
- Self-hosted coding / agents
- Distillation and research stacks
- API-compatible OpenAI-style clients
Strengths
Qualitative snapshot for architects—not a public ranking.
- Reasoning (R1)★★★★★
- Open weights★★★★★
- Price / performance★★★★★
- Coding★★★★★
- Enterprise packaging★★☆☆☆
Quick facts
- Founded
- 2023
- Headquarters
- Hangzhou, China
- Ownership
- Private
- Open source
- Yes
- Enterprise
- Yes
- Flagship model
- DeepSeek V4
On DataAIHub
- 3Models
- 3Products
- 3Tools
- 2GitHub
- 2Research
- 9Guides
- 13Benchmarks
Company profile
- Founded
- 2023
- Headquarters
- Hangzhou, China
- Founders
- Liang Wenfeng
- CEO
- Liang Wenfeng
- Funding
- Private
- Ownership
- Private
- Country
- China
- Primary focus
- Open-weight foundation models, Reasoning models, Efficient MoE training, Developer APIs
- Target users
- Developers, Researchers, Cost-sensitive enterprises
- Revenue model
- API usage, Open-weight ecosystem adoption
- Deployment
- Hosted API and chat plus open-weight self-hosting via vLLM/Ollama
- Licensing
- Open-weight model releases with separate API commercial terms
- Open source
- Yes
- Cloud provider
- No
- Website
- https://www.deepseek.com
- Confidence
- High
- Source coverage
- 14
Ecosystem
Competes with
- OpenAI
R1-level reasoning and coding force re-evaluation of closed API pricing.
- Anthropic
Competes for reasoning-heavy workloads that previously defaulted to Claude.
- Mistral AI
Open-weight peers for self-host and cost-sensitive production LLMs.
Works with
- vLLM
Default high-throughput serving path for DeepSeek open weights.
- Ollama
Fast local path for trying DeepSeek models.
- DeepSeek V4
Current API generation (Pro/Flash) for general and agent workloads.
Recommended for
- DeepSeek Models
How V4 / V3 / R1 fit coding, reasoning, and self-host stacks.
- Large Language Models
DeepSeek is a key open-weight reference point in today’s LLM landscape.
Often paired with
- Hugging Face
DeepSeek weights and discussion concentrate on the Hub ecosystem.
How DeepSeek evolved
Key moments in chronological order.
- Model
V4-Flash-Vision-Exp multimodal API
Experimental deepseek-v4-flash-vision-exp adds image+text at Flash rates; Files API enables free image reuse by file_id. API-only; open Flash weights remain text-first.
- API
V4 peak/off-peak API pricing live
Peak/off-peak rates replace the prior flat V4 API prices from 16:00 UTC. Peak hours 01:00–04:00 and 06:00–10:00 UTC; off-peak is half of peak.
- Model
DeepSeek V4-Pro generally available
V4-Pro-0813 GA on app/web/API with agent upgrades; peak/off-peak API prices from 2026-08-16 16:00 UTC.
- Open source
DeepSeek Harness developer preview
MIT-licensed agent harness (dsh) with a plugin architecture (Cordis). GitHub repo created 2026-08-13; preview APIs may break.
- Model
V4-Pro and V4-Flash bring 1M context and thinking modes to the DeepSeek API.
- Milestone
Industry cost/performance shock
R1 coverage forces hyperscalers and labs to revisit pricing and open-weight strategy.
- Model
Reasoning model resets expectations for open-weight chain-of-thought performance.
- Model
MoE open-weight model delivers frontier-competitive quality at lower cost.
- API
DeepSeek API platform
OpenAI-compatible API makes DeepSeek easy to try in existing clients.
- Founded
DeepSeek founded
Hangzhou lab focuses on efficient, high-capability open models.
Products
Foundation models
Related tools
Related research
- DeepSeek-V3 technical report
Original Paper
- DeepSeek-R1 technical report
Original Paper
GitHub
Related benchmarks
Related rankings
Related guides
Explore more companies
- AlibabaChinaGlobal technology group whose Qwen team releases competitive multilingual and multimodal foundation models for cloud and open-weight use.
- MetaUnited StatesMeta advances open multimodal research and releases the Llama open-weight model family used widely in self-hosted AI stacks. Muse Spark 1.3 (with Muse Code) is Meta’s paid Meta Model API path for agentic coding and multimodal agents. Combined with PyTorch, Llama still lets organizations train, fine-tune, and serve models without depending on a proprietary API.
- Mistral AIFranceMistral AI builds Mistral Large 3, Small/Medium commercial tiers, and open-weight models, plus La Plateforme APIs and Le Chat. Mixtral remains an Apache-2.0 self-host option; it is not listed on Mistral serverless API pricing as of 2026-08-16. Regional Endpoints (EU or US) are generally available.
- Moonshot AIChinaMoonshot AI builds the Kimi product family and releases large open-weight models. Kimi K3 (2.8T MoE, 1M context) is its flagship open multimodal agentic release under the Kimi K3 License, with API access at platform.kimi.ai and hosted access on Databricks Unity AI Gateway (2026-08-06).
- THUDM / TsinghuaChinaTsinghua University Knowledge Engineering Group (THUDM) — research lab behind LongBench and other influential open LLM evaluation and model work.
- AI21 LabsIsraelFoundation model company known for Jurassic and Jamba models, plus enterprise generative AI products for writing and document workflows.