Release 5.17.0
Release v5.17.0 New Model additions HYV4 Hy4-Preview is a 780B-parameter mixture-of-experts language model that activates 49B parameters per token. Each MoE layer holds 256 routed experts plus one always-active shared expert and routes every token to 8 of them. The context window is 1M tokens. The architecture combines four features: Multi-head Latent Attention (MLA) compresses keys and values into a low-rank latent ( kv_lora_rank ) that kv_b_proj expands back to one key/value per query head. De
阅读原始来源
来源记录
研究分析
变化记录
来源描述：Release v5.17.0 New Model additions HYV4 Hy4-Preview is a 780B-parameter mixture-of-experts language model that activates 49B parameters per token. Each MoE layer holds 256 routed experts plus one always-active shared expert and routes every token to 8 of them. The context window is 1M tokens. The architecture combines four features: Multi-head Latent Attention (MLA) compresses keys and values into a low-rank latent ( kv_lora_rank ) that kv_b_proj expands back to one key/value per query head. De
版本与成本证据
可供比较的近期同类记录：“Show HN: GC2 (Geo Compute Cloud)” (Show HN, 2026年11月7日)；“Sequoia shutting down its Scouts program marks the end of an era” (TechCrunch, 2026年11月7日)。主题相近仅用于导航，未核验其是否属于同一需求或独立证据。
适用与受影响对象
先明确变化影响的现有工作流，再判断是否升级。发布热度不证明效果提升；接口、价格或可用性变化也可能增加迁移工作。
缺失的比较信息
待补齐：新旧行为、受影响套餐与地区、生效时间、兼容性、回退路径及同一工作量下的成本。
采用与迁移验证
保存发布或政策原文，对新旧方案运行同一个任务并比较结果与完整账单；若涉及停服，先核验截止时间再安排迁移。