LLM Reference
DeepSeek

DeepSeek

61 models across 13 families · Latest: DeepSeek V4 Flash (2026-04)

Advancing artificial general intelligence (AGI).

CodingRAGAgentsLong contextVisionClassificationJSON / Tool useHighlight

DeepSeek's portfolio covers 60 active models across 13 current families, spanning coding, rag, and agents. Open a model detail page to compare provider routes and sourced benchmarks.

Covers 7 workload areas across 60 active tracked models; last verified 2026-06-24.

Use it for

  • Teams evaluating coding, rag, and agents across this lab's releases
  • Comparing model families before committing to a flagship
  • Migration and pricing follow-ups across 60 tracked models

Do not use it for

  • Choosing a hosting provider without opening a model page for price ladders

Active models

60

Current models from this lab, excluding deprecated ones

Active families

13

Current model families from this lab

Open catalog

60 open

23 open source / 37 open weights

Lowest output price

$0.030 /1M

Cheapest tracked output across active models, per 1M tokens

Latest dated release

2026-04-24

DeepSeek V4 Flash

Freshness

2026-06-24

Researched 8d ago

fresh

Information

Founded2023
Hangzhou, Zhejiang, China

Release cadence

Showing 5 recent dated releases (full timeline below). Latest: DeepSeek V4 Flash (2026-04-24).

Where this lab wins

  • Coding: 23 tracked models with SWE-bench / HumanEval-style scores.
  • RAG: 12 tracked models with ruler / needle retrieval benchmarks.
  • Agentic: 10 tracked models with BFCL, tau-bench, and SWE-bench tool-use coverage.
  • Long-context: 29 tracked models with context-token or InfiniteBench-class signal.

Flagship quality / price signal

Flagship: DeepSeek V3 (best sourced coding quality-per-dollar in this portfolio).

Coding task grade B · humaneval score 85.5 · cheapest tracked output $0.280 per 1M tokens

DeepSeek is a Chinese AI research organization founded in 2023. Advancing artificial general intelligence (AGI). DeepSeek ships 13 model families totaling 61 models, with the most recent release DeepSeek V4 Flash in 2026-04. Notable families include DeepSeek V4, DeepSeek OCR, and DeepSeek V3. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are added. View official API endpoints, benchmark performance, and coding/agent fit for every DeepSeek model.

About

DeepSeek is a Chinese AI company at the forefront of advancing artificial general intelligence (AGI). Their work involves developing AI models and tools for coding and mathematical tasks, such as DeepSeek Coder V2, an open-source coding model that supports over 300 programming languages and excels in both coding and math tasks. DeepSeek's research on AGI has the potential to revolutionize various industries beyond programming and math, such as healthcare, where AGI could assist in complex medical diagnoses and treatment planning, or finance, where it could optimize investment strategies and risk management. Software engineers and SaaS executives can follow DeepSeek's progress in AGI development to stay informed about the latest breakthroughs and potential applications of this transformative technology.

Featured models

ModelReleasedContextInput price ($/1M)Output price ($/1M)LicenseOpenness
DeepSeek V4 Flash2026-04-241m$0.14$0.28MITOpen source
DeepSeek V4 Pro2026-04-241m$0.435$0.87MITOpen source
DeepSeek OCR 22026-01-288k$0.03$0.03MITOpen source

Model families

Recent releases

  1. DeepSeek V4 Flash- 2026-04-24
  2. DeepSeek V4 Pro- 2026-04-24
  3. DeepSeek OCR 2- 2026-01-28
  4. DeepSeek V3.2- 2025-12-01
  5. DeepSeek OCR- 2025-10-21

Top comparisons

FAQ

Who founded DeepSeek and when?

DeepSeek was founded in 2023 and is associated with Hangzhou, Zhejiang, China.

What models has DeepSeek released?

DeepSeek ships 61 models across 13 families: DeepSeek V4, DeepSeek OCR, and DeepSeek V3.

Is DeepSeek's technology open source?

All tracked DeepSeek models are open-weight or open-source.

Where is DeepSeek headquartered?

DeepSeek is headquartered in Hangzhou, Zhejiang, China.

What is DeepSeek known for?

Advancing artificial general intelligence (AGI). Its most prominent tracked family is DeepSeek V4.

How can I access DeepSeek's models?

DeepSeek's models are available via DeepSeek Platform, Arcee AI, Alibaba Cloud PAI-EAS, AWS Bedrock, and Bitdeer AI.

Explore related pages

Last reviewed: 2026-06-24. Data sourced from public lab announcements and provider documentation.