awesome-ai-api

The world's largest open-source hub for AI API gateways & reseller reviews.
Curated · community-driven · fully transparent · every gateway probed daily.

🇨🇳 中文 · English · 🌐 Live site

---

What makes us different

We don't just list gateways — we probe every one of them daily for a real OpenAI-compatible /v1/models endpoint. No 200 OK on the homepage inflation.

• 🔌 Ground-truth API detection — every 🔌 badge = endpoint confirmed live
• 🔍 Transparency — no paid placement; raw data is Git-history-tracked
• 🌏 Global — bilingual (EN/中文), covers Western + Chinese markets
• 🛠 Developer-first — structured JSON, daily history snapshots, PR-editable
• 🤝 Community-driven — pricing & reviews by people who actually pay for these services

---

🎯 Quick Picks

Confused by 200+ gateways? Start here.

Use case	Recommendation	Why
🌐 Global, one API for all models	OpenRouter	Pioneer of the pattern; most stable, highest fees
👨‍💻 Best Claude Code in China	PackyAPI	Claude-Code native, active upstream, ¥1/sample
🏢 Enterprise / invoices	柏拉图AI · 云雾	Long-running, multi-region, invoices, DeepSeek/MidJourney too
💰 Cheap Claude (accept risk)	Terminal.Pub · XcodeBest	~0.15% of official; new = verify before topping up
🆓 Free trial credit	AnyRouter · 发现AI	Credits on signup, no card
🏠 Self-hosted	One API · New API	OSS; the engine 60% of the CN list runs on
⚡ Inference-only (Llama/Qwen/DeepSeek)	Groq · Together · Fireworks	Not resellers; own GPUs; US cards only
🌯 Multi-modal (video/music/image)	神马AI · 302.AI	MidJourney, Kling, Suno, PPT generators

⚠️ Never pre-pay large amounts. This industry has weekly rug-pulls. Top up ¥10 first, validate quality, then consider going up.

---

💚 Support this project

This repo is free, ad-free, and has no paid placement. If the leaderboard saved you from topping up a rug-pull, consider chipping in:

• USDT (any EVM chain — ERC20 / BSC / Polygon / Arbitrum / Base / Optimism):
  0xa5c74e7D3f0c8f1c0d7A395A6B7861Ab0A64cA7F
• ⭐ Star the repo — also helps a lot; boosts us in GitHub trending.
• 📝 Open a PR with a gateway report — the most valuable contribution.

💡 Double-check the chain before sending. Sending from an exchange? Withdraw on ERC20 / BSC / Polygon / Arbitrum / Base / Optimism. Do NOT use TRC20 (TRON) — that address doesn't exist on Tron.

---

Table of Contents

• 🏆 Full Leaderboard
• ❓ FAQ
• 📊 Categories
• 📝 Reviews
• ⚠️ Blacklist
• 🧪 Benchmark
• 🤝 Contributing
• 📜 License

---

🏆 Full Leaderboard

Auto-generated daily at 10:00 SGT (UTC+8) from live probes. See methodology · history · raw data.

🔌 = confirmed /v1/models endpoint (real API, not just a marketing page).

📊 Click to expand the full leaderboard

Last updated: 2026-09-24 15:12 (SGT)

Total: 189 gateways · 🔌 90 with confirmed /v1/models endpoint · 🟢 119 Verified · 🟡 6 Probable · 🧰 4 OSS · 🔍 60 Needs review

Top engines detected: new-api × 37 · one-api × 15 · dify × 3 · openrouter × 2 · litellm × 2

#	Gateway	Region	API	Models	Engine	Payment	Score	Latency	Tier
🥇	BUZZ	cn	🔌	claude, gpt, gemini	one-api	card	9.9	103 ms	🟢 Verified
🥈	Avian	global	🔌	13 models	—	—	9.9	168 ms	🟢 Verified
🥉	PackyAPI (PackyCode)	cn	🔌	claude, gpt, gemini	—	wechat	9.9	199 ms	🟢 Verified
4	AIHubMix	global	🔌	416 models	one-api	—	9.9	654 ms	🟢 Verified
5	呆呆兽中转站｜Claude、Codex、GLM、Kimi 折扣 API	cn	🔌	claude, gpt, anthropic	—	alipay, wechat, card	9.9	1101 ms	🟢 Verified
6	APIMart (apimart.ai)	global	🔌	claude, gpt, gemini	one-api	wechat, card	9.9	1390 ms	🟢 Verified
7	Featherless	global	🔌	gpt, openai, deepseek	one-api	—	9.8	157 ms	🟢 Verified
8	ePhone AI	cn	🔌	claude, gpt, gemini	—	—	9.8	203 ms	🟢 Verified
9	apipro.maynor1024.live	global	🔌	claude, gpt, gemini	—	—	9.8	251 ms	🟢 Verified
10	api.vveai.com	cn	🔌	claude, gpt, gemini	—	—	9.8	292 ms	🟢 Verified
11	api.gpt.ge	cn	🔌	claude, gpt, gemini	—	—	9.8	299 ms	🟢 Verified
12	api.v36.cm	cn	🔌	claude, gpt, gemini	—	—	9.8	322 ms	🟢 Verified
13	api.v3.cm	cn	🔌	claude, gpt, gemini	—	—	9.8	329 ms	🟢 Verified
14	Yuegle API	cn	🔌	claude, gpt, gemini	—	—	9.8	392 ms	🟢 Verified
15	AnyRouter · Universal AI Model Router	global	🔌	claude, gpt, gemini	openrouter	—	9.8	488 ms	🟢 Verified
16	TiMi CC	cn	🔌	gpt, openai, qwen	—	wechat, card	9.8	520 ms	🟢 Verified
17	jeniya.cn	cn	🔌	claude, gpt, chatgpt	—	—	9.8	1486 ms	🟢 Verified
18	ggwk1.online	cn	🔌	claude, gpt, gemini	new-api	—	9.7	126 ms	🟢 Verified
19	api-gptgod-work	global	🔌	322 models	—	—	9.7	626 ms	🟢 Verified
20	api-chatfire-cn	cn	🔌	claude, gemini, openai	—	—	9.6	128 ms	🟢 Verified
21	api.onechats.top	cn	🔌	claude, gemini, openai	new-api	—	9.6	201 ms	🟢 Verified
22	new.yunai.link	cn	🔌	claude, gemini, openai	—	—	9.6	233 ms	🟢 Verified
23	api.cursorai.art	global	🔌	claude, gpt, chatgpt	—	—	9.6	254 ms	🟢 Verified
24	api.soruxgpt.com	cn	🔌	claude, gemini, openai	new-api	—	9.6	289 ms	🟢 Verified
25	Aiberm	global	🔌	claude, gemini, openai	—	—	9.6	299 ms	🟢 Verified
26	api-deerapi-com	cn	🔌	claude, gemini, openai	—	—	9.6	343 ms	🟢 Verified
27	柏拉图AI_API中转站 (api.bltcy.ai)	cn	🔌	claude, gpt, openai	—	—	9.6	365 ms	🟢 Verified
28	api-mnapi-com	cn	🔌	claude, gemini, openai	—	—	9.6	479 ms	🟢 Verified
29	api.aipaibox.com	cn	🔌	claude, gemini, openai	new-api	—	9.6	1225 ms	🟢 Verified
30	LingxiCode	cn	🔌	claude, gemini, openai	new-api	—	9.6	1225 ms	🟢 Verified
31	marting.pro	global	🔌	gpt, chatgpt	—	—	9.5	188 ms	🟢 Verified
32	Doro AI	cn	🔌	claude, gpt, gemini	new-api	—	9.5	508 ms	🟢 Verified
33	comfly	global	🔌	gpt, openai	—	—	9.5	1166 ms	🟢 Verified
34	便携AI聚合API	cn	🔌	claude, gemini, openai	new-api	—	9.5	3123 ms	🟢 Verified
35	API Management	cn	🔌	openai	—	—	9.4	64 ms	🟢 Verified
36	zen-ai.top	cn	🔌	openai	—	—	9.4	84 ms	🟢 Verified
37	go.sbgpt.site	cn	🔌	openai	new-api	—	9.4	222 ms	🟢 Verified
38	钱多多 API (aigcbest.top)	cn	🔌	openai	new-api	—	9.4	276 ms	🟢 Verified
39	GPTNB ONEAPI	cn	🔌	gpt	—	—	9.4	318 ms	🟢 Verified
40	钱多多 API (api.ifopen.ai)	cn	🔌	openai	new-api	—	9.4	348 ms	🟢 Verified
41	钱多多 API (api.aigcbest.top)	cn	🔌	openai	new-api	—	9.4	526 ms	🟢 Verified
42	Unified API	cn	🔌	openai	—	—	9.4	761 ms	🟢 Verified
43	api-featherless-ai	global	🔌	—	—	—	9.3	64 ms	🟢 Verified
44	chatapi.onechats.top	global	🔌	—	new-api	—	9.3	116 ms	🟢 Verified
45	api.nekoapi.com	global	🔌	—	new-api	—	9.3	160 ms	🟢 Verified
46	zerocode.sbs	global	🔌	—	new-api	—	9.3	165 ms	🟢 Verified
47	api-gptsapi-net	global	🔌	—	—	—	9.3	177 ms	🟢 Verified
48	Loading...	global	🔌	—	—	—	9.3	179 ms	🟢 Verified
49	404 Not Found	global	🔌	—	—	—	9.3	231 ms	🟢 Verified
50	KKSJ API	global	🔌	—	new-api	—	9.3	239 ms	🟢 Verified

Top 50 shown. See data/_leaderboard.md for the full list of 189 gateways.

📌 Want your gateway listed? Open a PR with a filled gateway template. We accept any provider that meets our listing criteria.

---

❓ FAQ

What exactly is an "AI API gateway" or "中转站"?

A third-party service that exposes Claude, GPT, Gemini, DeepSeek, Qwen, and other frontier LLMs through a single OpenAI-compatible HTTP API. Most of them sit in front of the official providers and resell capacity, often at discounted prices, with friendlier payment options (WeChat/Alipay), or with extra features (rate-limit pooling, routing, logging, invoices).

How is this different from OpenRouter's /models page, helpaio.com, or apicompare.best?

• OpenRouter only lists gateways running inside OpenRouter's own routing graph. We list every relay we can find, including the dozens of small CN players.
• helpaio.com is a blog with human-written reviews. We publish machine-verified data (live /v1/models probes, engine fingerprints, uptime) alongside reviews.
• apicompare.best compares prices. We compare reachability, authenticity, and stability — cheaper doesn't help if the gateway disappears next week.

How do you decide the leaderboard score?

We combine: reachability (HTTP 2xx), confirmed /v1/models endpoint, real-model count, model-keyword density on the landing page, latency, and 30-day uptime from our daily snapshots. The formula lives in scripts/generate_leaderboard.py — no black box.

Is this a paid directory? Who pays you to be listed?

No one. There is zero paid placement. Listings are either auto-discovered or PR'd by the community. The repo is MIT-licensed and ad-free. If the leaderboard helped you, consider supporting the project.

Can I trust the top-ranked gateways with large amounts of money?

No. This industry sees weekly rug-pulls. A high score means a gateway is currently reachable, has a real API, and has been stable for 30 days — not that it is solvent or will still be online next month. Always top up small amounts first (≤ US$2), validate model quality, and scale slowly.

I got scammed by a gateway. How do I report it?

Open a GitHub Discussion under the Reports category with public evidence (screenshots, transaction IDs, Telegram group links). Two independent reports + a 14-day operator response window = the gateway lands on data/blacklist.json.

How do I add my gateway?

Add the URL to data/candidates.txt, then (optionally) add a human-curated name, region, and verdict to data/sites.yaml. Open a PR. The daily cron will pick it up.

Do you support non-OpenAI-compatible APIs (raw Anthropic, raw Gemini)?

We primarily probe /v1/models (OpenAI-compatible), because it's the lingua franca. Gateways that only expose raw Anthropic or Gemini endpoints can still be listed via data/sites.yaml with a verdict: likely_relay override and a note explaining the protocol.

---

📊 Categories

By Target Market

• 🌍 Global Gateways — OpenRouter, Together AI, Groq, Fireworks
• 🇨🇳 China Gateways — Road2All, DeepBricks, AiHubMix, OhMyGPT
• 🏠 Self-Hosted — One API, NewAPI, LiteLLM

By Specialty

• 🤖 Claude Code Compatible — Services fully supporting Claude Code CLI
• ⚡ Low-Latency — Sub-500ms TTFT
• 💰 Best-Price — Discounted vs. official rates
• 🔒 Privacy-First — No data retention / on-prem

By Payment

• 💳 Alipay/WeChat — For Chinese users
• 💳 Credit Card — International
• ₿ Crypto — Anonymous payment support

---

📝 Reviews

Deep-dive reviews and benchmark reports:

• 📊 2026 Q2 Gateway Benchmark (coming soon)
• 🔬 Claude Code Gateway Shootout (coming soon)
• 💸 Price Analysis: Official vs Gateway (coming soon)
• 🧪 Methodology

---

⚠️ Blacklist

Gateways that have been flagged for serious issues (rug-pulls, data leaks, fake models).

See blacklist.md for the current list and evidence archive.

If you've been affected, please open a GitHub Discussion with evidence. We review reports transparently.

---

🧪 Benchmark

We run periodic benchmarks testing:

• Latency (TTFT, total response time)
• Throughput (tokens/sec)
• Reliability (uptime %)
• Accuracy (model behavior matches official?)
• Pricing consistency (quoted vs. billed)

See scripts/benchmark.py for the test harness. Results in data/benchmarks/.

---

🤝 Contributing

We welcome PRs from users, gateway operators, and researchers. See CONTRIBUTING.md.

Quick ways to help:

• ✏️ Update a gateway's pricing or model list
• 📝 Submit a new gateway (use the template)
• 🚨 Report a scam or outage
• 🧪 Submit benchmark data
• 🌐 Improve translations

---

📜 License

Content is released under the MIT License. Attribution appreciated but not required.

---

🙏 Acknowledgments

Inspired by apicompare.best, helpaio.com/transit, and trustmrr.com — each taught us one piece of the puzzle: price data, content depth, and verified trust. We combine them with GitHub-native openness.

---

Built by Mack Ding and contributors · Made for the global OPC community

---

🦞 OPC Ecosystem

Built by @MackDing — One-Person Company infrastructure powered by AI agents.

Project	What it does
opc.ren	OPC founder hub — tools, signals, community
CodexClaw	Telegram bot for remote Codex access with MCP + subagent routing
awesome-ai-api	Leaderboard of 200+ AI API gateways & relays
claude-context-health	Diagnose & fix Claude Code session degradation
opc-daily-signal	AI-powered daily decision intelligence for OPC founders
doc-preprocess-hub	Enterprise document preprocessing — MinerU + docling