广告
加载中

智谱新一代模型提价10%,市场跟涨14%

胡镤心 2026-04-08 11:12
胡镤心 2026/04/08 11:12

邦小白快读

EN
全文速览

智谱发布新一代开源模型GLM-5.1并提价10%,带来显著市场反响和性能提升。

1. 模型能力:GLM-5.1是GLM-5的升级版,提升编程与推理能力,能持续工作8小时,在向量数据库优化任务中将查询性能从3500 QPS提升至21500 QPS,约6倍增长。

2. 价格变化:API输出价格从2.30美元涨至约2.53美元/百万Token,Coding Plan起价每月10美元。

3. 市场数据:股价上涨14.12%至890港元/股,市值达3964亿港元;2026年第一季度提价83%后调用量同比增长400%。

智谱通过模型升级和提价策略强化品牌营销,反映消费趋势和用户行为变化。

1. 品牌营销策略:逆势提价10%,基于模型能力提升,从过去降价竞争转向“量价齐升”,2025年API业务收入增长292.6%至1.90亿元。

2. 消费趋势:用户行为显示调用量不降反增,2026Q1同比增长400%,表明高端AI模型需求增长。

3. 产品研发启示:完全基于华为昇腾910B芯片训练,未用英伟达GPU,突显国产技术优势;模型参数7440亿,训练数据28.5万亿Token,提升长工程任务能力。

智谱的提价事件揭示增长机会和风险,需关注政策影响和新商业模式。

1. 政策解读与机会:API价格上调10%,但调用量增长400%,显示市场接受度高;MaaS API平台ARR达17亿元,过去12个月增长60倍,提示AI服务增长市场。

2. 风险提示:算力瓶颈可能导致服务质量下降,实际需求是当前算力支持的1至2倍,若无法缓解将影响用户留存。

3. 新合作方式:开源模型GLM-5.1发布,提供编程与推理增强能力,卖家可学习其优化任务性能6倍的经验。

智谱模型升级提供产品生产和数字化启示,突出商业机会。

1. 生产需求:模型完全基于华为昇腾910B芯片训练,替代英伟达GPU,提示芯片供应链机会;参数7440亿,MoE架构,每个Token激活约400-440亿参数,需高效硬件支持。

2. 商业机会:模型性能提升(如查询性能6倍增长)可应用于工业自动化;2025年总收入增长131.9%至7.24亿元,显示AI服务市场扩张。

3. 数字化启示:GLM-5.1能持续工作8小时,优化长工程任务,启示工厂推进AI在生产线监控等场景的应用。

行业趋势显示AI模型技术革新,但客户痛点突出,需提供解决方案。

1. 行业趋势:从降价竞争到逆势涨价,模型能力提升驱动;2026年API调用量增长400%,ARR达17亿元,增长60倍,表明MaaS服务需求激增。

2. 新技术:GLM-5.1增强编程与推理能力,上下文窗口200K Token,训练数据28.5万亿Token,提升长任务性能至21500 QPS。

3. 客户痛点与方案:算力紧张导致需求超出供给1至2倍,服务商可借鉴智谱优化方案(如600轮迭代提升性能),解决查询效率问题。

平台需应对商业需求和运营风险,智谱案例提供风向规避启示。

1. 平台需求与做法:MaaS API平台调用量增长400%后提价10%,显示价格管理策略;ARR达17亿元,增长60倍,提示招商机会。

2. 运营管理:模型基于华为芯片训练,避免依赖英伟达,启示平台硬件选择;服务需维持质量,算力瓶颈可能威胁留存。

3. 风险规避:提价后调用反增,但算力紧张是隐患,平台需监控供给,学习智谱优化任务性能6倍的经验以稳定服务。

产业动向揭示涨价逻辑和新问题,商业模式提供政策建议。

1. 产业动向:从烧钱降价到逆势涨价,模型能力提升驱动“量价齐升”;2025年API收入增长292.6%,2026Q1提价83%后调用量增400%。

2. 新问题:算力瓶颈导致需求超出供给1至2倍,服务质量维持成关键;训练完全国产化(华为芯片),凸显技术依赖风险。

3. 商业模式启示:开源模型GLM-5.1提供长工程任务能力(8小时持续工作),MaaS平台ARR达17亿元,增长60倍,建议政策支持算力扩容以保障可持续性。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

Zhipu AI has launched its new-generation open-source model GLM-5.1 and raised prices by 10%, generating significant market response and performance improvements.

1. Model Capabilities: GLM-5.1 is an upgraded version of GLM-5, enhancing programming and reasoning abilities. It can operate continuously for 8 hours and improves query performance from 3,500 QPS to 21,500 QPS in vector database optimization tasks—a roughly 6x increase.

2. Pricing Changes: API output pricing increased from $2.30 to approximately $2.53 per million tokens, while the Coding Plan starts at $10 per month.

3. Market Data: The stock price rose 14.12% to HK$890 per share, bringing market capitalization to HK$396.4 billion. After an 83% price hike in Q1 2026, API calls grew 400% year-over-year.

Zhipu AI strengthens its brand marketing through model upgrades and price increases, reflecting shifting consumption trends and user behavior.

1. Brand Marketing Strategy: Defying market trends with a 10% price hike, the move is based on improved model capabilities—shifting from price competition to a "volume and price rise" strategy. API business revenue grew 292.6% to RMB190 million in 2025.

2. Consumption Trends: User behavior shows API calls increased by 400% in Q1 2026 despite higher prices, indicating growing demand for high-end AI models.

3. R&D Insights: The model was trained entirely on Huawei's Ascend 910B chips without NVIDIA GPUs, highlighting domestic technological prowess. With 744 billion parameters and trained on 28.5 trillion tokens, it enhances long engineering task capabilities.

Zhipu's price increase reveals growth opportunities and risks, emphasizing the need to monitor policy impacts and new business models.

1. Policy Interpretation & Opportunities: Despite a 10% API price hike, call volume grew 400%, indicating strong market acceptance. The MaaS API platform reached RMB1.7 billion in ARR, growing 60x over the past 12 months, signaling expansion in AI services.

2. Risk Alert: Computing bottlenecks may degrade service quality, with actual demand exceeding current capacity by 1-2x. Failure to address this could impact user retention.

3. New Collaboration Models: The open-source GLM-5.1 model offers enhanced programming and reasoning capabilities. Sellers can learn from its 6x performance optimization in task efficiency.

Zhipu's model upgrade offers insights for production and digital transformation, highlighting commercial opportunities.

1. Production Demand: The model was fully trained on Huawei Ascend 910B chips, replacing NVIDIA GPUs—signaling supply chain opportunities. With 744 billion parameters and MoE architecture activating 40-44 billion parameters per token, efficient hardware support is critical.

2. Commercial Opportunities: Performance improvements (e.g., 6x query speed) can be applied to industrial automation. Total revenue grew 131.9% to RMB724 million in 2025, reflecting AI service market expansion.

3. Digital Insights: GLM-5.1’s 8-hour continuous operation and long-task optimization inspire factories to adopt AI for production line monitoring.

Industry trends show AI model innovation, but customer pain points require solutions.

1. Industry Trends: The shift from price cuts to strategic price hikes is driven by model upgrades. API calls grew 400% in 2026, with ARR reaching RMB1.7 billion—a 60x increase—indicating surging MaaS demand.

2. New Technology: GLM-5.1 enhances programming and reasoning with a 200K-token context window and 28.5 trillion training tokens, boosting long-task performance to 21,500 QPS.

3. Customer Pain Points & Solutions: Computing shortages cause demand to exceed supply by 1-2x. Service providers can adopt Zhipu’s optimization methods (e.g., 600 training iterations) to improve query efficiency.

Platforms must address commercial demands and operational risks, with Zhipu’s case offering risk mitigation insights.

1. Platform Needs & Strategies: A 10% price hike after 400% call growth demonstrates effective pricing management. ARR of RMB1.7 billion (60x growth) highlights partnership opportunities.

2. Operational Management: Training on Huawei chips reduces reliance on NVIDIA, informing platform hardware choices. Service quality must be maintained amid computing bottlenecks threatening user retention.

3. Risk Mitigation: While calls rose post-price hike, computing constraints remain a risk. Platforms should monitor supply and learn from Zhipu’s 6x task optimization to stabilize services.

Industry trends reveal pricing logic and emerging challenges, with business models informing policy recommendations.

1. Industry Dynamics: The shift from subsidy-driven pricing to strategic hikes reflects model-driven "volume-price synergy." API revenue grew 292.6% in 2025, with calls rising 400% after an 83% price increase in Q1 2026.

2. Emerging Issues: Computing bottlenecks cause demand to exceed supply by 1-2x, challenging service quality. Full domestic training (Huawei chips) highlights technology dependency risks.

3. Business Model Insights: Open-source GLM-5.1 supports long engineering tasks (8-hour operation). MaaS platform ARR of RMB1.7 billion (60x growth) suggests policy support for computing expansion to ensure sustainability.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

【亿邦原创】4月8日,智谱发布新一代开源模型GLM-5.1。据模型聚合平台OpenRouter显示,伴随此次发布,智谱GLM再度提价10%。截至发稿,智谱股价报890港元/股,上涨14.12%,市值3964亿港元。

GLM-5.1是GLM-5的增量升级版本,通过增强的后训练流程重点提升了编程与推理能力。基础架构和GLM-5一致,总参数量7440亿,MoE架构,每个Token激活约400-440亿参数,上下文窗口200K Token,训练数据量28.5万亿Token。

GLM-5.1的更新在于长工程任务能力。当前大模型以分钟级交互为主,而GLM-5.1能够在单次任务中持续、自主工作长达8小时。智谱官方博客披露,GLM-5.1在向量数据库优化任务中,历经超过600轮迭代、调用超6000次工具,最终将查询性能从3500 QPS提升至21,500 QPS,约为原先的6倍。

此外,GLM-5.1完全基于华为昇腾910B芯片训练,未使用任何英伟达GPU。

伴随此次发布,智谱GLM再度提价10%。在OpenRouter上,GLM-5.1的价格信息没有直接显示,但GLM-5的API定价为输入0.72美元/百万token,输出2.30美元/百万token。Coding Plan的起价是每月10美元。据此推算,GLM-5.1提价10%后,输出价格约从2.30美元涨至2.53美元/百万Token。

提价背后,是智谱2025年财报验证的“量价齐升”逻辑。

2025年,智谱总收入7.24亿元,同比增长131.9%。其中开放平台及API业务收入1.90亿元,同比增幅高达292.6%。截至2026年3月,MaaS API平台ARR已达17亿元,过去12个月增长60倍。2026年第一季度API调用价格累计上调83%后,调用量不降反增,同比增长400%。

一年前,国产大模型厂商还在以降价90%以上的方式争夺市场份额。从“烧钱换增长”到“逆势涨价”,这背后既有模型能力提升,也有算计供给紧张。智谱在财报电话会中多次坦承,实际需求约为当前算力支持的1至2倍。若算力瓶颈无法缓解,涨价后的服务质量能否维持,将直接影响用户留存。

文章来源:亿邦动力

广告
微信
朋友圈

这么好看,分享一下?

朋友圈 分享

APP内打开

+1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0