https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's tech blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, distributed training, and data engineering. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Sat, 11 Jul 2026 00:00:00 GMT https://www.youngju.dev/blog/2026-07-11-tencent-hy3-open-model.en https://www.youngju.dev/blog/2026-07-11-tencent-hy3-open-model.en On July 6, 2026, Tencent released Hy3 under Apache 2.0: a Mixture-of-Experts reasoning and agent model with 295B total parameters but only 21B active per token. Here is what is genuinely new, where it sits in the Chinese open-weight wave, and how far to trust the vendor-reported benchmark numbers. Sat, 11 Jul 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) hy3tencenthunyuanopen-weightsmoellm https://www.youngju.dev/blog/2026-07-11-tencent-hy3-open-model.ja https://www.youngju.dev/blog/2026-07-11-tencent-hy3-open-model.ja Tencent は2026年7月6日、Hy3 を Apache 2.0 で公開しました。総計295Bのうち21Bだけが活性化する MoE の推論・エージェント言語モデルです。何が実際に新しいのか、中国系オープンウェイトの流れの中でどこに位置するのか、そしてベンダーが自ら出したベンチマーク数値をどこまで信じるべきかを整理します。 Sat, 11 Jul 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) hy3tencenthunyuanopen-weightsmoellm https://www.youngju.dev/blog/2026-07-11-tencent-hy3-open-model https://www.youngju.dev/blog/2026-07-11-tencent-hy3-open-model Tencent이 2026년 7월 6일 Hy3를 Apache 2.0으로 공개했다. 총 295B 중 21B만 활성화되는 MoE 추론·에이전트 언어 모델이다. 무엇이 실제로 새로운지, 중국계 오픈 웨이트 흐름에서 어디에 놓이는지, 그리고 벤더가 직접 낸 벤치마크 숫자를 어디까지 믿어야 하는지 정리했다. Sat, 11 Jul 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) hy3tencenthunyuanopen-weightsmoellm