SaaS vs Private
Side-by-side comparison · Decision guidance · Billing and data ownership differences
Same features — the differences are multi-tenant structure, billing model, and model autonomy. This page helps you decide.
SaaS or private?
Answer 6 questions to get a recommendation
- 1
Must data live inside your company?
- 2
Need to go live in 1–2 weeks?
- 3
Need to use domestic / self-trained models without public APIs?
- 4
Subject to MLPS-3 / finance / healthcare compliance?
- 5
Do you have a dedicated IT/ops team?
- 6
Expecting to scale from pilot to prod within 6 months?
60-Second Decision Table
| Your key constraint is… | Pick |
|---|---|
| Data must stay on-premise | Private |
| Need to start using it within 1–2 weeks | SaaS |
| Need to use domestic / self-trained models (no public API) | Private |
| Budget-first, don't want to operate it | SaaS |
| High compliance/audit requirements (MLPS Level 3+) | Private |
| Want to pilot fast and migrate later | Start with SaaS, migrate to Private later |
Full Comparison
| Dimension | SaaS Cloud | Private Deployment |
|---|---|---|
| Tenant hierarchy | Organization → Workspace (two-layer) | Workspace (single layer · 1 deployment = 1 organization) |
| Model access | Mainstream models provided by the Evose platform | Customer-configured models + tool services |
| Custom models | Not supported | Supported (LLM / image / audio / video) |
| Billing | Credit-based | No platform billing (customer pays model costs directly) |
| Data ownership | Managed by platform (isolated + encrypted) | Owned by customer |
| Deployment & ops | Run by the Evose team | Run by customer's IT |
| Onboarding time | Instant activation | Standard 2–4 weeks (5 phases) |
| Updates & upgrades | Auto-updated | Follow the product release cadence (customer-controlled) |
| Best for | Small to mid-size teams · Fast pilots · Elastic needs | Mid to large enterprises · Heavy compliance · Sensitive data |
Shared Capabilities (identical across both)
Regardless of which deployment you choose, the following capabilities are completely identical — no functionality is lost when migrating:
- Agent Builder · Workflow · Chatflow · Knowledge Base
- Workbench
- RBAC + ACL permission governance
- Workspace resource isolation
- Three pillars of observability (Logs / Metrics / Traces)
- Highly available LLM service (Round Robin + Failover)
- Multi-channel deployment (web / iOS / Android / DingTalk / WeCom / Feishu / WeChat)
Billing Models in Detail
SaaS: Credits
- Unified billing unit that abstracts pricing across providers
- Consumption dimensions: model calls · tool calls · knowledge base storage · data source capacity
- Attribution: Credits are attributed at three layers — workspace / user / resource (per Agent / Workflow)
- Budget control: organization-level spend caps + usage alerts
Private: No Platform Billing
- The customer pays model provider costs directly (OpenAI / Anthropic / on-prem GPU)
- Evose takes no revenue share
- Consumption can still be attributed inside Evose (for internal cost allocation), but is not uploaded to the platform
Private Deployment: Standard 5-Phase Rollout (2–4 weeks)
→ Detailed Private Deployment Guide
A Hybrid Adoption Suggestion
If you have both "fast pilot" and "future private deployment" requirements:
- Months 0–3: Start on SaaS, validate 2–3 core scenarios (customer service / marketing / R&D assistant)
- Months 3–6: Accumulate users and data, confirm ROI
- Months 6–9: Launch private deployment, migrate smoothly from SaaS (the product is identical; the cost is the data migration itself)
Next Steps
- Picked SaaS → SaaS 5-minute walkthrough
- Picked Private → System Requirements → Docker Deployment
- Still comparing → Security Overview