通过统一 API 接入头部供应商的最新 AI 模型，价格透明，用多少付多少。
在任意受支持的供应商上使用你自己的密钥 —— 平台费 0%
按任务对比成本 →成本计算器
Anthropic flagship
Anthropic balanced
Google fast multimodal
Zhipu flagship
Fast and efficient Claude model
Previous generation powerful Claude model
Anthropic Claude Sonnet 4.5 — balanced speed/quality
DeepSeek V4 Flash — fast OpenAI-compatible model with automatic context caching and an optional thinking mode (reasoning tokens counted inside completion tokens).
DeepSeek V4 Pro — higher-capability OpenAI-compatible model with automatic context caching and an optional thinking mode.
Google Gemini 2.5 Flash — fast reasoning model with a thinking budget. Billed output includes thinking (thoughts) tokens.
Z.ai GLM-4.5-Air — lighter, cost-efficient OpenAI-compatible model with a 128K context window and automatic context caching.
Z.ai GLM-4.6 — flagship OpenAI-compatible model with a 200K context window, optional thinking mode (reasoning tokens counted inside completion tokens), and automatic context caching.
Zhipu
OpenAI (relay param integration pending)
Moonshot Kimi K3 — long-context OpenAI-compatible model with a 1M context window and automatic context caching (requires temperature=1 or omitted).