腾讯混元发布带视觉的 Hy4,自称稳居开源模型第一梯队
cihangxie · x · 2026-11-04
Tencent Hunyuan officially released Hy4, adding vision capabilities and claiming it is "firmly among the leading open models," with gains across finance, frontend, game, officework and research scenarios, and noting it was built responsibly with cyber/biosecurity in mind.
Team member Yonglong Tian shared that folks still love high-performing open-weight models and said the team is taking a cautious approach to releasing the weights, inviting users to try it out.
所属事件:腾讯混元发布旗舰模型 Hy4 正式版:新增视觉与电脑操作(29 条相关)→
「模型」频道最新
- Top 30 模型性价比图谱:GPT-6.1 Sol 成价值王,Opus 5.5 最贵 — daniel_mac8 · 2026-11-05
- Seoul Meetup 透露 MiniMax 将推新模型 M3 Pro — ollama · 2026-11-05
- 用户观点:宁要 1M 上下文的 27B 稠密模型,也不要 128k 的大 MoE — Express_Quail_1493 · 2026-11-05
- Wes Roth 深度解读:OpenAI 报告模型安全任务暗中摆烂与权重空间编程 — Wes Roth · 2026-11-05
- Kimi K3.1 开源不到 24 小时,社区已推出 uncensored 版 — rrwilliams · 2026-11-05
- 小互预感 OpenAI 今晚发新模型:多位用户体感模型"变聪明了" — xiaohu · 2026-11-05