模力方舟 MoArk (Gitee AI)
OSChina's model API and compute market, built around Chinese AI chips
Overview
MoArk (模力方舟), formerly Gitee AI, is the model platform run for OSChina, the company behind Gitee. It offers one OpenAI- and Anthropic-compatible API to 200+ models paid from prepaid token packs, hourly GPU rental including Chinese accelerators, fine-tuning, and an app marketplace with billing APIs.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Online model plaza with the free-trial access token
- Function calling
- 1M context
- Cache support
- Multimodal
- 1M context
- Cache support
- 128K and 1M context tags
- Cache support
- Multimodal
- 128K and 1M context tags
- Cache support
- 购买全模型 Token 资源包
- Bulk, high-concurrency purchases unlock special prices
What you can produce with 模力方舟 MoArk (Gitee AI)
- OpenAI- and Anthropic-compatible API to 200+ models with automatic failover
- Prepaid all-model token packs, plus 100 free trial calls per user per day
- Dedicated model deployment
- Hourly GPU rental, including MetaX and Enflame accelerators, with SSH or JupyterLab access
- Fine-tuning for Qwen text models and FLUX and Qwen Image models
- An AI app marketplace with user-auth and billing APIs
- Integration guides for Claude Code, Dify, n8n and ComfyUI
MoArk is Gitee AI under a new name, and its pitch is now enterprise token supply rather than a model playground. Prices match its rivals (Qwen3.8-Flash at ¥0.8/¥2.7 per million tokens, DeepSeek-V4.1-Flash at ¥2/¥8), paid from a prepaid all-model pack, with 100 free trial calls per user per day.
What genuinely sets it apart is hardware: models run on, and compute is rented from, MetaX, Enflame, Biren, Iluvatar and Ascend accelerators as well as NVIDIA, which matters to Chinese institutions with domestic-chip requirements. The 420,000 enterprise users figure is self-reported. Everything is in Chinese, so for teams outside China it is a niche choice.
What we found
模力方舟 MoArk is the model platform that launched in 2023 as Gitee AI and took its Chinese name in February 2025; the site, now at moark.com, still shows both names. It is operated by 深圳奥思研工智能科技有限公司 for OSChina (开源中国), the company behind the Gitee code-hosting service and the OSChina developer community, and that existing audience is its main distribution advantage.
The product has moved from a model playground toward what it now calls an enterprise token-management platform. One OpenAI- and Anthropic-compatible endpoint reaches 205 models, with optional automatic failover to another compute pool when one fails. Prices are listed per million tokens: Qwen3.8-Flash costs ¥0.8 in and ¥2.7 out, DeepSeek-V4.1-Flash ¥2 and ¥8, GLM-5.3 ¥8 and ¥28, and Qwen3.8-Max ¥12 and ¥36. Usage is paid from a prepaid all-model resource pack of any amount, with bulk discounts by negotiation. For trying models, the FAQ offers 100 free trial calls per user per day in the online model plaza; earlier versions of this page said the allowance was per model, and that no allowance was shown.
What separates MoArk from PPIO or SiliconFlow is hardware. Its model plaza groups models by the accelerator they run on: MetaX, Enflame, Biren, Iluvatar CoreX and Huawei Ascend clusters alongside NVIDIA. Its compute market rents MetaX and Enflame GPUs by the hour with SSH or JupyterLab access. That matters to Chinese institutions with domestic-hardware requirements, and much less to anyone else. The platform also offers fine-tuning for Qwen text models and FLUX-style image models, an app marketplace with user-login and billing APIs for paid AI apps, and setup guides for Claude Code, Dify and n8n.
MoArk reports 420,000 enterprise users and 2,500 university users; both figures are self-reported and undated. Third-party reviews outside Chinese developer forums are scarce. The documentation, console and payment flow are Chinese-only, and nothing on the site addresses Southeast Asian data residency, so this is a domestic-market product.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including 模力方舟 MoArk (Gitee AI)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with 模力方舟 MoArk (Gitee AI) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to 模力方舟 MoArk (Gitee AI) directly →
Spotted something out of date? Suggest an update →
Alternatives to 模力方舟 MoArk (Gitee AI)
More in LLMs & Chat
