Minimum two users. Monthly billing is USD 25/user. Usage limits and additional credits can apply; an API integration has separate billing. Model and tool availability varies by workspace.
ข้อมูลและการดำเนินงาน
Business content is not used to train models by default. This is cloud processing: review retention, connectors and the applicable data-processing terms.
Two to 150 members; monthly billing is USD 25/seat. Standard seats have usage limits. Do not infer unlimited agent or coding work, or included API usage.
ข้อมูลและการดำเนินงาน
Commercial data is not used for model training by default. Connected services, retention and access permissions still need review.
Reference base price excludes temporary promotions. This is a suite subscription, not a standalone model API. Check feature limits, supported languages and country checkout.
ข้อมูลและการดำเนินงาน
Google says covered Workspace data is not used for generative AI training outside your domain without permission. Check which account and service employees use; personal accounts and integrations can have different terms.
พิจารณาหากงานของธุรกิจอยู่ใน Microsoft 365 เป็นหลัก
เหตุผล หลักฐาน และข้อจำกัด
ข้อจำกัดที่ต้องตรวจสอบ
A qualifying Microsoft 365 licence costs extra; up to 300 users. Uses ordinary annual list price, excluding the first-year offer. Check tenant, market, app and agent entitlements.
ข้อมูลและการดำเนินงาน
Microsoft states covered prompts, responses and Microsoft Graph data are not used to train foundation models. Existing file permissions matter; review oversharing before enabling access.
Exact model: gemma3:4b. Current Ollama package is about 3.4 GB; working memory and context need more room. Smaller download is not evidence of better answers or suitability for complex decisions.
ข้อมูลและการดำเนินงาน
Open weights under Gemma Terms, not Apache 2.0. Check permitted use and downstream obligations. A local runtime alone does not make external tools or cloud services private.
Exact model: qwen3.5:9b. Ollama lists different runtime builds (about 6.6–7.6 GB at review). Download size is not required RAM. Test memory, latency and local-language accuracy on your actual tasks.
ข้อมูลและการดำเนินงาน
Apache 2.0 model licence. Local inference can keep prompts on your hardware, but cloud models, web tools, remote embeddings and integrations can send data elsewhere.
สำหรับทีมที่ต้องการสร้างผู้ช่วยความรู้หรือแอป AI บนคลาวด์
เหตุผล หลักฐาน และข้อจำกัด
ข้อจำกัดที่ต้องตรวจสอบ
Three team members and 5,000 monthly model message credits at the reviewed tier. Model choice changes credit consumption; bringing your own API key creates separate model costs. Taxes excluded.
ข้อมูลและการดำเนินงาน
Your documents and workflow data use hosted infrastructure and selected model providers. Check retention, deployment region and document access controls before uploading business material.
Dify licence adds conditions to Apache 2.0, including multi-tenant and frontend branding restrictions. Server minimums exclude local model inference; model quality and infrastructure sizing need separate tests.
ข้อมูลและการดำเนินงาน
Self-hosting gives deployment control, not automatic offline operation. Verify models, embeddings, plugins, logs and integrations. A cloud model still processes the data you send to it.
2,500 workflow executions per month at this tier. Execution counts are not model tokens. Check AI allowances, connected service charges, retries and human approvals before unattended use.
ข้อมูลและการดำเนินงาน
Workflows are hosted by n8n; the chosen model and connected apps introduce additional processors. Review logs, credentials and hosting region.
Source-available Sustainable Use Licence permits internal business use with conditions; resale is a separate question. Paid features are not included. The AI starter kit needs production hardening.
ข้อมูลและการดำเนินงาน
You run the server. Configure every model, database and tool deliberately; a self-hosted workflow can still send data to cloud APIs. Budget for patching, backups, monitoring and access control.
intent.asia