WPS 365
Kingsoft's enterprise office AI platform (金山办公 WPS 365). AI runs inside the document, and Qingzhou AI puts private deployment on a single CPU server. From 199 RMB per seat a year.
Overview
Kingsoft Office has been shipping WPS since 1988, a year before Microsoft Office existed, and WPS 365 is the enterprise repackaging of that suite with AI built in. It bundles Writer, Spreadsheets, Presentation, multi-dimensional tables, meetings, enterprise mail and collaboration into one tenant, sold per seat at 199, 399 or 599 RMB per user per year. The AI features live only in the 599 RMB flagship tier with a five-seat minimum, which is worth knowing before anyone budgets from the headline price.
The product decision that makes it technically interesting arrived on 24 April 2026 with Qingzhou AI. Rather than assume a GPU cluster, Kingsoft rebuilt the private deployment path around CPU-served auxiliary models, a monolithic architecture in place of microservices, one unified document parsing and vectorisation base and a single prompt management plane. The claimed result is that a proof of concept that previously needed ten servers and eight high-end accelerators now runs on one standard CPU box, a reduction of over 95% in hardware footprint. For Chinese state-owned enterprises and government bodies, which face both a hard prohibition on sending documents to public APIs and a hard scarcity of accelerator allocation, that is the constraint that actually binds.
The Enterprise Brain layer, launched into Asia-Pacific on 9 July 2026, is the retrieval and integration tier: it indexes internal documents and hooks into existing systems so the AI operates inside the open contract or spreadsheet rather than in a separate chat pane. Kingsoft's disclosed case study describes a technology company cutting document review from 140 minutes to 40 across 400-plus audit types. Take vendor case studies at their stated weight. What is verifiable is that WPS 365 pulled 720 million RMB in 2025, up 64.93% year on year, so the enterprise motion is past the pilot stage. For an engineer, the honest assessment is that this is procurement and integration work, not a build target. There is no real developer platform here, English documentation is thin and the CPU-served models trade capability for deployability by design.
Key Features
- ✓ Per-seat subscription at 199, 399 and 599 RMB per user per year, with WPS AI available only in the 599 RMB flagship tier and a five-seat floor
- ✓ Qingzhou AI (轻舟AI), announced 24 April 2026, a private-deployment architecture that runs on one standard CPU server instead of the ten servers and eight accelerators the previous on-premises build required, cutting hardware footprint by over 95%
- ✓ AI operating inside the open file rather than in a side chat window, so contract review, spreadsheet analysis and slide drafting act on the document itself
- ✓ Enterprise Brain layer, taken to Asia-Pacific on 9 July 2026, indexing company documents and connecting to existing internal systems
- ✓ One tenant covering Writer, Spreadsheets, Presentation, multi-dimensional tables (多维表格), meetings, enterprise mail and collaboration
- ✓ Architectural choices aimed squarely at on-premises: CPU-served auxiliary models, a monolith instead of microservices, a unified parsing and vectorisation base and a single prompt management plane
Where it holds
- • The CPU-only private deployment path is the genuine differentiator. Removing the GPU requirement changes which organisations can run on-premises AI at all
- • Kingsoft sits on domestic-substitution (信创) procurement lists, so in state-owned enterprise and government tenders it is frequently the only compliant option on the table
- • OOXML format fidelity with Microsoft Office is the best in the Chinese market, which is the practical reason large organisations can migrate without a document rewrite project
- • Enterprise traction is measurable rather than aspirational: 720 million RMB in WPS 365 revenue for 2025, growing 64.93% year on year
Where it breaks
- • WPS AI ships only in the 599 RMB per seat flagship tier with a five-seat minimum, so the 199 and 399 RMB tiers buy collaboration with no AI at all
- • Models small enough to serve on CPU are small enough to notice. Long-context contract reasoning and complex multi-document synthesis will not match a frontier model on a GPU endpoint
- • English product documentation and the international admin console are considerably thinner than the Chinese equivalents, and the Asia-Pacific rollout only began in July 2026
- • There is no meaningful API or extension platform for engineers. This is a suite you deploy, integrate and configure rather than one you build applications on top of
My Take
Almost every enterprise AI pitch in China assumes you have a GPU cluster. Kingsoft went the other direction in April 2026 with Qingzhou AI, a private-deployment build that serves auxiliary models on CPU and fits on one standard server, against the ten servers and eight accelerators the previous on-premises architecture wanted. For a state-owned enterprise that cannot send documents to a public API and cannot get an accelerator allocation either, that is the whole argument. Price it honestly though: WPS AI appears only in the 599 RMB per seat per year flagship tier, five seats minimum, and models small enough to run on CPU are small enough that you will feel the ceiling.
Quick Info
- Pricing:
- freemium
- Openness:
- Proprietary
- Starting at:
- Four tiers on a per-seat annual subscription: a free 体验版 trial tier, 商业协作版 at 199 RMB per user per year, 商业高级版 at 399 RMB per user per year and 商业旗舰版 at 599 RMB per user per year. WPS AI is gated to the 599 RMB flagship tier only, with a five-seat minimum purchase. Qingzhou AI private deployment is sold separately and flexibly, either as a per-account subscription scaled to headcount or as a component-based licence bought by the unit. WPS 365 revenue reached 720 million RMB in 2025, up 64.93% year on year.
- Added:
- Aug 2026
- Updated:
- Aug 2026
Use Cases
Judge it on your own work
The notes above say where WPS 365 holds and where it breaks. The fastest check is your own workload.
Visit website ↗Alternatives to WPS 365
Writer
paidEnterprise AI content and agent platform on Writer's own Palmyra models, with a knowledge graph, brand guardrails and audit trails. No free tier, Starter is $29 per user a month annually.
Lark CLI
open-sourceMIT-licensed official CLI for Lark and Feishu (飞书), from ByteDance. 200+ commands and 24 agent skills over Docs, Base, Sheets, Calendar and approvals, built for AI agents.