이 이벤트의 기반 데이터
2-
AI 모델
DeepSeek-V4
The efficiency school scales up: 1.6T parameters, MIT license, 1M context.
DeepSeek-V4 went live in preview on April 24, 2026 in Pro and Flash variants with API access and open weights under MIT. V4 Pro is a 1.6-trillion-parameter mixture-of-experts activating 49B per token, pairing MoE with sparse attention for a 1M-token context, and posting 80.6% on SWE-bench Verified (as reported in launch coverage), frontier-class coding from…
레코드 열기 → -
기업
DeepSeek
The lab that shook the market: frontier results at a fraction of the compute.
Spun from quant fund High-Flyer by Liang Wenfeng, DeepSeek trained V3 on a 2,048-H800 cluster for a disclosed ~$5.6M final-run cost, then released the R1 reasoning model openly in January 2025, triggering a ~$600B one-day fall in NVIDIA's market value as investors re-priced compute efficiency. Its open-weights MoE releases remain the benchmark for…
레코드 열기 →