ShiyunApi ⚠️ Discontinued Review: Pricing & Comparison
⚠️ Discontinued as of June 2026; the official site now redirects to TokenRiver (tokenriver.cn)
Last verified: 2026-06-23 · Visit official site →
ShiyunApi overview
ShiyunApi’s main selling point was “native support for all three protocols” — meaning it didn’t just build an OpenAI compatibility layer, but natively supported Anthropic’s and Gemini’s own protocol formats as well, which was friendlier for developers using each vendor’s official SDK directly (like the anthropic Python SDK or google-generativeai) who didn’t want to change their code. Multiple third-party reviews showed it had integrated 650+ models (including the GPT-5.x, Claude 4.x, and Gemini 3.x lineups), with overall pricing roughly 8-9.5% below official rates. The official site claimed 99.99% SLA, time-to-first-token in the 20-30ms range, and a full enterprise management system (sub-accounts, fine-grained key management, compliant invoicing), positioned for enterprise customers.
Discontinuation notice and migration guidance
ShiyunApi discontinued operations in June 2026. The former shiyunapi.cn domain now redirects to TokenRiver (tokenriver.cn), with the brand and team carrying over.
If you previously used a ShiyunApi API key or had it integrated into your business, here’s a recommended migration path:
- Confirm the base URL: TokenRiver’s onboarding docs provide a new
base_url— just replace your old ShiyunApi endpoint - Three-protocol compatibility carries over: TokenRiver inherited the three-protocol design (OpenAI/Anthropic/Gemini native protocols), so existing SDK code usually needs little modification
- Re-issue your key: your old API key may have been invalidated after the shutdown — sign up on TokenRiver and get a new one
- Verify your model list: after migrating, confirm the models you need (e.g., Claude 4.x, Gemini 3.x lineups) are available on the new platform
Information verified 2026-06-23. See the latest official TokenRiver announcements for discontinuation and migration details.
Successor platform: TokenRiver
The destination ShiyunApi’s site now redirects to, tokenriver.cn, is an upgraded enterprise-grade product — evolved from a personal/developer relay tool into an “enterprise-grade large-model intelligent gateway platform.” TokenRiver’s main features:
Domestic routing, data stays in-country: the service is deployed domestically, with the entire call chain never routing through overseas nodes, meeting data-compliance requirements.
High availability: officially commits to 99.995% SLA, with multi-channel disaster recovery and automatic failover.
Model coverage: TokenRiver’s models come from major cloud-vendor channels (Tencent Cloud, Alibaba Cloud, Baidu Cloud, Volcano Engine) and direct official supply (Zhipu GLM, DeepSeek, Kimi, MiniMax, Qwen, Doubao), leaning toward mainstream domestic large models. If your use case centers on domestic models like DeepSeek, Qwen, or GLM, TokenRiver’s coverage is more complete than an overseas-focused relay.
Endpoint: https://api.tokenriver.cn, interface format compatible with the OpenAI SDK.
New-user credit: sign up and get a million tokens of free credit, ready to try without topping up first.
Enterprise management features
TokenRiver offers a set of enterprise-oriented management features that didn’t exist in the ShiyunApi era:
- Security governance: sensitive-word filtering, audit logs, PII data masking, meeting enterprise compliance review needs
- Cost control: multi-dimensional quota limits, real-time expense tracking, with detailed billing records queryable per call
- Organization management: supports Feishu/WeCom SSO integration and department-level permission isolation, suited to multi-team collaboration
- Observability: real-time call-chain monitoring, performance metrics tracking
Related reviews
- Tencent Cloud Hunyuan: Tencent’s official Hunyuan model, deeply integrated with the Tencent Cloud ecosystem
- 302.AI: pay-as-you-go with zero monthly fee, supports 100+ models
- iFlytek Spark: iFlytek’s official platform, strong on both speech and text, vertical strength in education and healthcare
- YunWu API: 500+ aggregated models, ¥0.5/USD exchange rate, free daily GPT-4o calls via GitHub login
Quick facts
| Pricing model | ⚠️ Discontinued — please migrate to TokenRiver (tokenriver.cn) |
|---|---|
| Model coverage | 650+ models, full coverage across all three protocols (OpenAI/Anthropic/Gemini native protocols), including the GPT-5.x/Claude 4.x/Gemini 3.x lineups |
| Latency / SLA | Claims 99.99% SLA; multiple reviews cite time-to-first-token in the 20-30ms range (vendor self-reported/third-party reviews, not independently verified) |
| Mainland direct connect | 直连 |
| Best for | Enterprise / Developers |
| Referral program | No public affiliate program found. |
Pros
- Native support for all three protocols (not just an OpenAI compatibility layer), friendlier to projects using the Anthropic/Gemini SDKs directly
- Claims 99.99% SLA and a full enterprise management system, suited to teams with high stability requirements
- Frequently ranks near the top for model coverage (650+) and overall score in industry comparison articles
Cons
- Positioned for enterprise use, pricing may exceed an individual developer's budget
- Figures like 99.99% SLA and time-to-first-token are mainly self-reported by the vendor or from vendor-affiliated review articles, with no independent third-party verification found
Compare more AI API relays
See the full comparison board — filter by price tier, model coverage, and mainland direct-connect status.
Back to the comparison board →