DeepSeek
Open-weight model family notable for efficient training at frontier scale.
Обзор
DeepSeek is an open-weight model family whose published training methods attracted attention for reaching competitive quality at lower reported cost. It solves the cost question in training rather than in inference, which is where most optimisation effort goes. It exists as evidence that architecture and training-schedule choices matter enough to change what a given budget can produce, and its weights and methods are published.
Граф технологий
Как DeepSeek связана с остальным стеком.
Использует и зависит от
- Transformer architecture Algorithms DeepSeek based on Transformer High ✓ source
Применяется к
- MoE architecture Algorithms MoE used for DeepSeek High ✓ source
Инструменты
ПО, которое реализует DeepSeek или интегрируется с ним.
Infrastructure
Альтернативы
Сравнения
Подробные сопоставления с альтернативами выше.