DeepSeek-V4 launch : millions of contexts, AI capabilities are further enhancedDeepSeek-V4 preview version officially released and open-sourced, supporting millions of contexts, with significant improvements in agent capabilities, knowledge coverage, and reasoning performance. This release includes two major versions: Pro and Flash. Among them, Pro offers stronger performance but is limited by high-end computing power, which is currently scarce. As new generation computing nodes are deployed, costs are expected to decrease significantly. This means AI capabilities are entering a new leap phase, and the limitation is no longer the model itself, but the allocation of computing power. When performance breakthroughs and computing bottlenecks occur simultaneously, the true opportunities often lie beneath the surface.

Voir l'original
post-image
Cette page peut inclure du contenu de tiers fourni à des fins d'information uniquement. Gate ne garantit ni l'exactitude ni la validité de ces contenus, n’endosse pas les opinions exprimées, et ne fournit aucun conseil financier ou professionnel à travers ces informations. Voir la section Avertissement pour plus de détails.
  • Récompense
  • Commentaire
  • Reposter
  • Partager
Commentaire
Ajouter un commentaire
Ajouter un commentaire
Aucun commentaire
  • Épingler