DeepSeek
Le laboratoire dont les modèles de raisonnement ouverts ont changé la donne
Le texte intégral est présenté en anglais ; le titre et le résumé sont localisés.
CE QUE C'EST
DeepSeek was founded in 2023, incubated by the quantitative fund High-Flyer. It focuses on foundation-model research, centred on mixture-of-experts architectures and reasoning models, and keeps releasing model weights. It shipped V3 in late 2024 and the reasoning model R1 in early 2025.
Pourquoi c'est important
R1 shipped near-frontier reasoning with open weights and a published training method, directly resetting expectations for how much compute reasoning models need and pushing peers to reprice.
Jalons clés
- 2023
Founded, incubated by High-Flyer
- 2024
Released DeepSeek-V3 with open weights
- 2025
Released the reasoning model DeepSeek-R1