Trust Center

Last updated: 2026-08-16

We do not serve entities registered in mainland China, or accounts operated primarily on their behalf. Individuals must be in jurisdictions allowed by each upstream model provider. See Compliance.

1. Company & contacts

  • Product: ModelAPI (aimodelapi.ai).
  • Support: support@aimodelapi.ai
  • Sales / contracts: sales@aimodelapi.ai
  • Security: security@aimodelapi.ai
  • Compute: AWS us-east-1 (Virginia) primary gateway and AWS ap-southeast-1 (Singapore) PoP. We do not run inference compute in mainland China.
  • Legal entity name and mailing address are provided on enterprise contracts. Email sales for a counterparty pack — we will not invent a public registered address here.

2. Upstream channel types

  • OEM / default IDs: licensed platform routing (not scraped consumer ChatGPT/Claude web sessions).
  • AKL-* and *-rsc: licensed ArkLin channel. RSC SKUs are discounted variants of the same family.
  • BR-* / Bedrock: AWS Bedrock when the SKU is listed and capacity is healthy. Bedrock IDs are not silently substituted onto default OEM names.
  • CN providers: DeepSeek, Alibaba (Qwen), Moonshot (Kimi), Zhipu (GLM) via provider APIs.
  • Media catalogs: IMG-*, SD-*, WAN-* for image / video / 3D.
  • Public listed chat/reasoning IDs you can call today: 58 (see /models). We do not advertise 400+ if the callable catalog is smaller.

3. Data retention & training

Prompts are routed to upstream providers to fulfill your request. We do not use API content to train ModelAPI models. Limited logs are kept for billing disputes, abuse investigation, and legal compliance.

提示词会转发至上游以完成推理。我们不用 API 内容训练自有模型。仅在计费争议、滥用调查和法律要求范围内有限保留。

4. How to verify a model

  • The JSON response includes a model field. Compare it to the ID you requested.
  • Channel prefixes (AKL-, BR-, IMG-) are part of the ID — they are not hidden aliases for a cheaper substitute.
  • Copy the curl on each model page and run it; check latency and the returned model id.
  • Daily endpoint probes are on /status. Model-level public probe history is being wired to the same page — treat seeded bars as illustrative until live probe series is published. Peak vs measured vs contract SLA: /performance.

5. Live regions, uptime, latency

Live today: Global (Primary gateway) (api.aimodelapi.ai), Singapore (sg.api.aimodelapi.ai). Planned (not billed as live): Virginia (planned), Tokyo (planned), Bangkok (planned), Frankfurt (planned), Alt Primary Domain (planned).

  • api.aimodelapi.ai — Global (Primary gateway) — live · published 90-day target 99.97%
  • us.api.aimodelapi.ai — Virginia — planned
  • sg.api.aimodelapi.ai — Singapore — live · published 90-day target 99.95%
  • jp.api.aimodelapi.ai — Tokyo — planned
  • th.api.aimodelapi.ai — Bangkok — planned
  • eu.api.aimodelapi.ai — Frankfurt — planned
  • api2.aimodelapi.ai — Alt Primary Domain — planned

P50/P95 are measured from the probe client on /status and /tools/api-latency-test. Gateway uptime targets are not a substitute for upstream model SLAs.

6. Security audit & SLA

  • Last public security review note: 2026-05-27 (practices on /security). A dated third-party pentest report will be linked here when issued.
  • Gateway availability target and support times: /sla. Upstream provider outages are excluded.
  • Credits: unused balances do not expire on accounts in good standing. Refunds follow /terms — prepaid credits are non-refundable except where required by law or explicitly granted.
  • If we terminate an account for cause, unused prepaid credits may be forfeited as stated in the Terms.

7. Status of this page

Live endpoints right now: api.aimodelapi.ai, sg.api.aimodelapi.ai. For the callable catalog see /models. For benchmarks see /benchmarks.