可用 Chat API
顶级大语言模型，Extended Thinking 推理，即刻接入生产环境。
Gemini 4 Argon
Gemini 4 Argon API on PoYo. Google DeepMind's frontier reasoning model with 1M unified context & output for long-horizon coding, knowledge work, and cybersecurity at $1.60 input and $8.00 output per 1M tokens (20% off official).
Claude Haiku 5.5
Anthropic’s compact model for fast classification, extraction, summaries, and subagent workflows with adaptive reasoning and a 1M-token context.
GPT-6.1 Sol
GPT-6.1 Sol API on PoYo. Use Chat Completions or Responses at $1.6 input and $8 output per million tokens, 20% below official standard rates.
Claude Sonnet 5.5
Claude Sonnet 5.5 API on PoYo. Use Chat Completions or Claude Messages at $1.6 input and $8 output per million tokens, 20% below official standard rates.
GPT-6 Luna
GPT-6 Luna API on PoYo. Use Chat Completions or Responses at $0.08 input and $0.4 output per million tokens, 20% below official standard rates.
GPT-6 Sol
GPT-6 Sol API on PoYo. Use Chat Completions or Responses at $1.6 input and $8 output per million tokens, 20% below official standard rates.
Claude Opus 5.5
Claude Opus 5.5 API on PoYo. Use Chat Completions or Claude Messages at $3.2 input and $16 output per million tokens, 20% below official standard rates.
Grok 4.7
Grok 4.7 API on PoYo. Use Chat Completions at $1.6 input and $4.8 output per million tokens, 20% below official standard rates.
DeepSeek V4.1 Flash
DeepSeek V4.1 Flash is a 552B MoE chat model with native vision, 1M context, 384K max output, and 8B/16B active parameters for cheaper agent workloads.
GPT-6 Astra
GPT-6 Astra API for advanced reasoning, coding, and agent workflows. Use Chat Completions or Responses with pay-as-you-go pricing on PoYo.
Claude Fable 5.1
Anthropic's most capable generally available model for long-horizon coding, knowledge work, research, vision, 1M context, and up to 128K output.
Gemini 3.8 Flash
Google's best reasoning and coding Flash model for long-horizon software engineering, autonomous agents, and enterprise workflows.
Grok 4.6
xAI's frontier model for long-running agents, coding, engineering, knowledge work, and ambitious interactive or visual projects.
Gemini 3.7 Flash
Google's GA Flash model for agentic coding, multimodal reasoning, web development, long-context knowledge work, and tool-using applications.
Claude Opus 5
Anthropic's flagship agentic model for advanced coding, long-horizon execution, computer use, deep research, and professional knowledge work.
Kimi K3
Moonshot AI's 2.8T-parameter flagship multimodal model with a 1M-token context window, built for long-horizon coding, deep reasoning, knowledge work, and tool-using agents.
GPT-5.6
GPT-5.6 API access for Sol, Terra, and Luna through the Responses API, from high-throughput automation to frontier coding and agent workflows.
Claude Sonnet 5
Claude Sonnet 5 API access for agentic coding, long-running agents, browser and computer use, professional workflows, 1M context, and up to 128K output.
Gemini 3.5 Flash
Gemini 3.5 Flash API access for chat, coding, reasoning, and production agent workflows.
Claude Opus 4.8
Claude Opus 4.8 supports long-context chat, agentic coding, professional reasoning, high-output workflows, 1M context, and up to 128K output.
DeepSeek V4 Flash
DeepSeek V4 Flash is a fast, economical DeepSeek V4 API model with 1M-context chat, 284B total parameters, and 13B active parameters.
DeepSeek V4 Pro
DeepSeek V4 Pro is a stronger reasoning, coding, and agent workflow model with 1M-context chat, 1.6T total parameters, and 49B active parameters.
Claude Opus 4.7
Anthropic's most capable generally available model for frontier coding, deep reasoning, and high-resolution vision.
GPT-5.5
GPT-5.5 API access for chat, coding, reasoning, and production agent workflows.
GPT-5.4
GPT-5.4 API access for chat, coding, reasoning, and production agent workflows.
GPT-5.2
GPT-5.2 API access for chat, coding, reasoning, and production agent workflows.
Claude 4.6 API
Claude Sonnet 4.6 and Claude Opus 4.6 API access for coding, agents, computer use, and long-context reasoning.
Gemini 3 Series
Google Gemini 3 系列 — Flash 和 Pro，1M 上下文窗口，动态思考模式。
Claude 4.5 Series
Anthropic Claude 4.5 系列 — Opus、Sonnet、Haiku，80.9% SWE-bench，Extended Thinking。
平台对比
poyo.ai vs OpenAI vs Anthropic vs Google
poyo.ai Chat API 与官方平台的生产级对比。
| 对比项 | poyo.ai | OpenAI | Anthropic | Google AI |
|---|---|---|---|---|
| 模型覆盖 | Claude + Gemini 全系列 | 仅 GPT 系列 | 仅 Claude 系列 | 仅 Gemini 系列 |
| 价格优势 | 透明积分制 | 官方定价 | 官方定价 | 官方定价 |
| 服务稳定性 | 生产级，稳定 QoS | 官方服务 | 官方服务 | 官方服务 |
| API 兼容性 | OpenAI SDK 兼容 + 原生格式 | OpenAI SDK | Anthropic SDK | Google AI SDK |
| 上下文窗口 | 最高 1M tokens | 最高 128K | 最高 200K | 最高 1M |
常见问题
常见问题解答
关于 AI Chat API 的常见问题。
目前支持 Claude 4.5 系列（Opus、Sonnet、Haiku）和 Gemini 3 系列（Flash Preview、Pro Preview），共 5 个顶级模型。
复杂推理和 Agent 开发选 Claude Opus 4.5；日常平衡任务选 Sonnet 4.5；高并发低成本选 Haiku 4.5；快速编码选 Gemini 3 Flash；深度分析选 Gemini 3 Pro。
完全兼容。/v1/chat/completions 端点支持 OpenAI SDK，只需更换 base URL 即可使用所有 5 个模型。
Claude 4.5 的深度推理模式，逐步展示思考过程。Gemini 3 则提供 Dynamic Thinking Levels，可按需配置推理深度。
Gemini 3 系列支持 1M tokens，Claude 4.5 系列支持 200K tokens，均支持最高 64K tokens 输出。
按输入/输出 tokens 消耗积分。积分永不过期，透明可预测，无订阅费用。