Rostered 2026-10-25 15:50 Beijing time · 10 lanes · 23 models on duty
Relay365 takes OpenAI-compatible, native Anthropic and native Gemini calls on a single key, and books every request on what it actually used.
| Model lane | Duty light | Wire it up via | Charged against | Input rate starts at, per 1M | Shift note |
|---|---|---|---|---|---|
| AnthropicClaude Fable 5 · Claude Fable 5.1 · Claude Haiku 4.5 · Claude Opus 5 · Claude Opus 5.5 · Claude Sonnet 5.5 | 6 models on duty | Anthropic Messages wire · /v1/messages | Wallet, or a Claude duty pass | $0.3542 | Official Claude Code or Claude Desktop only, version 2.1.285 and later |
| OpenAIGPT-5.6 Sol · GPT-6 Astra · GPT-6.1 Sol | 3 models on duty | OpenAI wire · /v1/chat/completions and /v1/responses | Wallet, or a GPT duty pass | $0.1486 | Codex and any OpenAI SDK patch straight in |
| GoogleGemini 2.5 Flash · Gemini 3.1 Pro · Gemini 3.8 Flash · Gemini 4 Argon | 4 models on duty | Gemini generateContent wire · /v1beta | Wallet only | $0.2389 | Gemini CLI takes the bare host; streams add alt=sse |
| DeepSeekDeepSeek V4 Flash · DeepSeek V4 Pro | 2 models on duty | OpenAI-compatible wire · /v1 | Wallet only | $0.2889 | Any OpenAI-compatible client can take the call |
| Qwen line · AlibabaQwen3.7 Plus · Qwen3.8 Max | 2 models on duty | OpenAI-compatible wire · /v1 | Wallet only | $0.15 | Any OpenAI-compatible client can take the call |
| Kimi line · MoonshotKimi K3 | 1 models on duty | OpenAI-compatible wire · /v1 | Wallet only | $1.17 | Any OpenAI-compatible client can take the call |
| GLM line · ZhipuGLM-5.3 | 1 models on duty | OpenAI-compatible wire · /v1 | Wallet only | $0.6472 | Any OpenAI-compatible client can take the call |
| Grok line · xAIGrok 4.6 · Grok 4.7 | 2 models on duty | OpenAI-compatible wire · /v1 | Wallet only | $0.5653 | Any OpenAI-compatible client can take the call |
| MiniMaxMiniMax M3 | 1 models on duty | OpenAI-compatible wire · /v1 | Wallet only | $0.1528 | Any OpenAI-compatible client can take the call |
| Hunyuan line · TencentHunyuan 4 | 1 models on duty | OpenAI-compatible wire · /v1 | Wallet only | $0.4097 | Any OpenAI-compatible client can take the call |
A green light only means the lane is on the current catalogue. It is not an uptime figure, and the desk never invents one.
Heavy Claude and GPT users can take a duty pass: quota is counted at official prices and the dedicated key draws only from it.
Claude models only; $700 of official-price quota, up to $380 per 7 days; drawn at official price x2.5; valid 14 days, no daily request cap.
Log inClaude models only; $1400 of official-price quota, up to $450 per 7 days; drawn at official price x2.5; valid 30 days, no daily request cap.
Log inClaude models only; $3200 of official-price quota, up to $900 per 7 days; drawn at official price x2.5; valid 30 days, no daily request cap.
Log inClaude models only; $6000 of official-price quota, no weekly ceiling; drawn at official price x2.5; valid 30 days, no daily request cap.
Log inevery GPT / Codex model; $3000 of official-price quota, up to $900 per 7 days; drawn at official price x2.5; valid 30 days, no daily request cap.
Log inevery GPT / Codex model; $7200 of official-price quota, up to $2100 per 7 days; drawn at official price x2.5; valid 30 days, no daily request cap.
Log inWhen a request ends, the desk prices the token counts upstream reports on four separate meters and adds them up.
When the wallet runs dry, or a duty pass reaches its quota or 7-day ceiling, the API answers HTTP 402 until you top up or add a pass.