Google Gemini Enterprise Agent Platform 定价体系公布
Google Gemini Enterprise Agent Platform Pricing Structure Announced
EVENT SUMMARY
事件摘要
Google 公布 Gemini Enterprise Agent Platform 完整定价体系，该平台取代旧版 AI Platform 和 AutoML，统一整合 AI 训练与推理服务。核心变化包括：不再支持低成本低性能机型推理和 scale-to-zero，新增 TensorFlow 运行时优化、模型共托管及 30 秒增量计费。Agent Runtime 和 Sandbox 环境中轮次间空闲等待时间不计费，使用量按秒取整。Skill Registry 于 2026 年 9 月 1 日起按存储和读写操作计费，费率与 Memory Bank 一致，Vulnerability Analysis 模型 Token 另计。Semantic Governance Policy（用于自然语言约束 Agent 工具调用）将于 2026 年晚些时候开始按评估 Token 计费。
ANALYST VIEW
中立分析
为什么重要
Google 通过统一平台整合 AI 训练与推理服务，取消 scale-to-zero 可能提高低流量场景成本门槛，但 30 秒增量计费和空闲时间不计费增强了短时任务和交互式 Agent 的成本竞争力。将 AI 治理策略（Semantic Governance Policy）和技能管理（Skill Registry）作为独立计费模块，体现了 Agent 能力模块化商业化的行业趋势。
影响对象
从旧版 AI Platform/AutoML 迁移的开发者需关注低成本机型不可用和 scale-to-zero 取消带来的成本变化；交互式 Agent 开发者受益于空闲时间不计费和按秒计费粒度；合规敏感型企业需评估 Semantic Governance Policy 的治理成本；维护大量可复用 Skills 的企业需关注存储与读写开销。
证据说明
所有信息均来自 Google 官方定价页面的产品对比说明及各组件计费规则描述，包括平台替代关系、计费粒度、生效时间等。
不确定性
旧版 AI Platform 和 AutoML 的最终退役时间表未明确；Semantic Governance Policy 具体生效日期仅标注为 2026 年晚些时候，评估模型 SKU 和费率未披露；Skill Registry 中 Vulnerability Analysis 使用的具体模型和费率未说明；Skill Registry 当前是否可用及是否有免费额度未明确。
SOURCES
1 个来源 · 4 条记录
以下记录均来自同一来源，不构成多来源相互印证。
来源原文：The costs for Agent Platform remain the same as they are for the legacy AI Platform and AutoML products that Agent Platform supersedes, with the following exceptions: Legacy Agent Platform Inference and AutoML Tables predictions supported lower-cost, lower-performance machine types that aren't supported. Legacy Agent Platform Inference supported scale-to-zero, which isn't supported. Agent Platform also offers more ways to optimize costs, such as Optimized TensorFlow runtime, Support for co-hosting models, No minimum usage duration for Training and Prediction — charged in 30 second increments.
来源原文：Runtime and Sandbox environments, including Code Execution and Computer Use, are billed on Agent Compute (vCPU-hours) and Agent Memory (GiB-hours) resource usage. All usage is rounded to the nearest second. For Runtime, idle time spent waiting for the next prompt between turns is not billed.
来源原文：Semantic Governance Policies are natural language constraints that you configure to govern AI agents through their tool calls. You are billed for agent-model response evaluations and evaluation-model tokens (used for agent-model response evaluations) under their respective model SKUs. Note: Semantic Governance Policy billing on this pricing structure will commence later in 2026.
来源原文：Skill Registry pricing is based on the following components. Storage: Total data stored, billed as Agent Storage ($0.30/GiB-month). Operations: Read API requests: 1 Agent Compute vCPU-h ($0.085) for every 3 million read operations. Write API requests: 1 Agent Compute vCPU-h ($0.085) for every 1 million write operations. Tokens: model tokens used for Vulnerability Analysis are billed separately. Note: Skills Registry billing will commence on September 1st, 2026.