APIKEY.FUN

💰 Pricing & Plans

🔧 API & Developer Experience

• •API Style: OpenAI-compatible /v1/chat/completions endpoint with Bearer-token auth (also accepts x-api-key and x-goog-api-key header formats) — existing OpenAI SDK clients work by swapping base URL and key only.
• •Base URL: https://apikey.fun/v1 — drop-in replacement for https://api.openai.com/v1, also serves /v1/models for live catalog discovery.
• •Model Catalog: ~50 models in one key: GPT-4o, Claude 3.5 Sonnet, DeepSeek-V3, Llama 3.3 70B, Qwen 2.5 72B, plus Midjourney for image generation — the only aggregator in this list that ships Midjourney behind the same endpoint as text LLMs.
• •Auth & Top-up: Email + password registration (whitelisted suffixes: qq.com / 163.com / gmail.com / *.edu.cn etc.); prepaid top-up via dashboard; API keys can be created with per-key usage caps for sub-account isolation.
• •Channels for Teams: Admin panel exposes group-based RPM multipliers and per-user rate-limit overrides — built-in for resellers / teams who want to allocate quotas to downstream users without standing up their own gateway.
• •2FA & Account Security: TOTP-based 2FA enabled platform-wide; passkey option in development. Login agreement modal with explicit consent timestamps logged per account revision (rev 72ca0174… as of 2026-05-20).
• •Streaming & Multimodal: SSE streaming for chat completions on all text models; image endpoints take natural-language prompts and return hosted image URLs (Midjourney). No native audio/video endpoints exposed as of 2026-08-31.

🎨 One Key, Mixed Text + Image (incl. Midjourney)

APIKEY.FUN's differentiator is bundling Midjourney image generation behind the same OpenAI-style endpoint that serves GPT-4o and Claude — a single Bearer token, one bill, no separate MJ subscription, no Discord relay bot. For a Chinese indie developer or small team this collapses three vendors (text LLM, image LLM, payment processor) into one prepaid account that accepts domestic-rail top-up. The catalog of ~50 models is narrower than OpenRouter's 300+, but every model listed is one a typical mainland team actually needs: GPT-4o for general reasoning, Claude 3.5 Sonnet for long-context code review, DeepSeek-V3 / Qwen 2.5 for cost-sensitive Chinese workloads, Llama 3.3 70B for open-weight parity, and Midjourney for visual content. The trade-off is vendor transparency: you don't see which upstream provider the platform is sourcing from, so debugging rate-limit or quality regressions requires going through APIKEY.FUN support rather than checking OpenAI / Anthropic status pages directly.

🌐 China Access & Latency

APIKEY.FUN's apikey.fun endpoint is one of the few aggregator endpoints on this list that is directly reachable from mainland China without proxy — the platform sits on domestic-friendly CDN / hosting and is positioned specifically for the China-to-overseas-model use case. Round-trip latency from a China Telecom / China Unicom / China Mobile line is typically 80-200 ms for short prompts, comparable to calling a domestic provider like DeepSeek or Kimi, and substantially better than the 250-500 ms that OpenRouter or portkey show when routed through public overseas nodes. There is no China-specific pricing tier — you pay the same per-token rate as overseas users, but the bill settles in CNY via domestic top-up channels (Alipay / WeChat / bank card) rather than requiring a foreign credit card, which removes the single biggest blocker for solo Chinese developers signing up. The downside is reliability variance: as a smaller aggregator, uptime is sensitive to upstream provider outages (when OpenAI or Anthropic degrade, you degrade with it, often with less graceful fallback than OpenRouter's auto-routing). For mission-critical production traffic, pair APIKEY.FUN with a direct DeepSeek or Qwen account as a hot-standby.