0

DeepSeek

Chinese AI lab spun out of the High-Flyer quant fund; shook the field in early 2025 with R1, an open-weights reasoning model trained at a fraction of frontier cost.

Visit
Type
open lab
HQ
Hangzhou, China
Founded
2023
Funding
Self-funded by High-Flyer Capital Management
Website
deepseek.com
GitHub
github.com/deepseek-ai
Twitter
twitter.com/deepseek_ai

Cite

Notes

Only stored in your browser.

Evals
0
Tools
0
Models
37
Papers
1
Boards
0
People
2

Models

37

Ranked by Artificial Analysis Intelligence Index

#ModelReleasedIntelligence
1DeepSeek V4 Flash 073131 Jul 202651.8
2DeepSeek V4 Flash Vision (Reasoning, Max Effort)21 Aug 202651.5
3DeepSeek V4 Pro 042324 Apr 202645.3
4DeepSeek V4 Flash 042324 Apr 202642.1
5DeepSeek V3.21 Dec 202525.1
6DeepSeek V3.2 Speciale1 Dec 202522.6
7DeepSeek V3.2 Exp29 Sep 202521.7
8DeepSeek V3.1 Terminus22 Sep 202521.7
9DeepSeek V3.121 Aug 202521.4
10R120 Jan 202518.6
11DeepSeek V3 032425 Mar 202515.2
12DeepSeek V326 Dec 202414.2
13R1 Distill Qwen 32B20 Jan 202511
14DeepSeek R1 0528 Qwen3 8B29 May 202510.3
15R1 Distill Llama 70B20 Jan 20259.8
16DeepSeek R1 Distill Qwen 14B20 Jan 20259.7
17DeepSeek-V2.5 (Dec '24)10 Dec 20246.5
18DeepSeek-V2.56 Sep 20246.4
19DeepSeek R1 Distill Llama 8B20 Jan 20256.2
20DeepSeek-Coder-V217 Jun 20244.7
21DeepSeek R1 Distill Qwen 1.5B20 Jan 20253.3
22DeepSeek-V2-Chat6 May 20243.3
23DeepSeek Coder V2 Lite Instruct17 Jun 20242.7
24DeepSeek LLM 67B Chat29 Nov 20232.6
25R1 052828 May 2025-
26DeepSeek R120 Jan 2025-
27DeepSeek R1-ZeroJan 2025-
28DeepSeek-V2-Lite-Chat15 May 2024-
29DeepSeek Reasoner--
30DeepSeek Coder--
31deepseek--
32DeepSeek R1t2 Chimera--
33DeepSeek Chat--
34DeepSeek Chat V3--
35DeepSeek Prover V2--
36DeepSeek V4 Pro Max--
37DeepSeek-V3.1-Think--
37 / 37 models

Papers

1

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

preprint

2025