Organization
DeepSeek
Tracked models
39
Largest training run
9.7 x 10^24 FLOP
39 tracked AI models list DeepSeek as a developer. The largest is DeepSeek-V4-Pro, at 9.7 x 10^24 FLOP (April 2026). The most recent tracked release is from April 2026.
It was last valued at $50B. It employs around 300 people.
Models (39)
DeepSeek-V4-Pro9.7 x 10^24 FLOPDeepSeek-V3.24.2 x 10^24 FLOPDeepSeek-V3.2-Exp4.2 x 10^24 FLOPDeepSeek-R1 (May 2025)4 x 10^24 FLOPDeepSeek-V3.1-Terminus3.6 x 10^24 FLOPDeepSeek-V3.13.6 x 10^24 FLOPDeepSeek-R13.5 x 10^24 FLOPDeepSeek-V3 (Mar 2025)3.3 x 10^24 FLOPDeepSeek-V33.3 x 10^24 FLOPDeepSeek-V4-Flash2.5 x 10^24 FLOPDeepSeek-V2.51.8 x 10^24 FLOPDeepSeek-Coder-V2 236B1.3 x 10^24 FLOPDeepSeek-V2 (MoE-236B)10^24 FLOPDeepSeek LLM 67B8 x 10^23 FLOPDeepSeek Coder 33B4 x 10^23 FLOPDeepSeek-VL-7B10^23 FLOPDeepSeekMath 7B10^23 FLOPDeepSeek LLM 7B8.4 x 10^22 FLOPDeepSeek Coder 6.7B8 x 10^22 FLOPDeepSeekMoE-16B3.4 x 10^22 FLOPDeepSeek Coder 1.3B1.6 x 10^22 FLOPDeepSeek-VL-1.3B8.7 x 10^21 FLOPDeepSeek-LLM-1.3b-base3.9 x 10^21 FLOPDeepSeekMath-V2November 2025Deepseek OCROctober 2025DeepSeek-Prover-V2-671BApril 2025DeepSeek-Prover-V2-7BApril 2025Janus-Pro-1BJanuary 2025Janus-Pro-7BJanuary 2025DeepSeek-R1-Distill-Llama-8BJanuary 2025DeepSeek-R1-Distill-Qwen-1.5BJanuary 2025DeepSeek-R1-Distill-Qwen-32BJanuary 2025DeepSeek-R1-Distill-Qwen-7BJanuary 2025DeepSeek-R1-Distill-Llama-70BJanuary 2025DeepSeek-R1-Distill-Qwen-14BJanuary 2025DeepSeek-R1-ZeroJanuary 2025DeepSeek-R1-Lite-PreviewNovember 2024Janus 1.3BOctober 2024DeepSeek-Prover-V1.5August 2024
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.
Other organizations
xAIOpenAIMeta AIGoogle DeepMindNVIDIACursorAnthropicByteDanceMoonshotAlibabaHuaweiSaudi AramcoInflection AIBeijing Academy of Artificial Intelligence / BAAIReka AIAI SingaporeGoogleGoogle ResearchMicrosoftChina TelecomZ.ai (Zhipu AI)Xiaomi CorpETH ZurichEcole Polytechnique Federale de Lausanne (EPFL)