Model List

Browse ready-to-use model IDs, default recommendations, and common use cases.

These are the model IDs currently available through the AI API Proxy. We recommend copying model names exactly as shown, including capitalization and hyphens. For multipliers, pricing, context windows, and image understanding support, see Model Pricing and Billing.

Default Recommended Model

claude-sonnet-4-6

Note: The AI API Proxy's recommended high-concurrency model. It has a 1M context window, supports image understanding, and is a strong first choice for agents, workflows, image understanding, long documents, and high-concurrency scripts.

claude-sonnet-4-6 is a high-concurrency aggregated model optimized by the AI API Proxy for response speed, token generation speed, a 1M context window, agent tool calls, and image understanding. It works well as a long-term default model in tools such as OpenClaw, Claude Code, Hermes, WorkBuddy, n8n, and Coze.

Global AI Models

OpenAI (Top Recommendation) 🌟🌟🌟🌟🌟

gpt-5.4

Note: OpenAI's primary general-purpose model. It has a 1M context window, supports image understanding, and is suitable for everyday Q&A, coding, long documents, and agent tasks.

gpt-5.5

Note: A highly capable OpenAI model. The platform currently lists a 258K context window and image understanding support. It is suitable for complex reasoning, code planning, high-quality writing, and mission-critical output.

gpt-5.3-codex-spark

Note: A real-time Codex coding model with a 128k context window. Its key advantage is speed, making it suitable for interactive coding, quick patches, UI refinements, and lightweight engineering tasks.

gpt-image-2

Note: A high-quality image generation and editing model billed per image. It is suitable for product images, posters, design assets, cover images, and creative work that requires higher visual quality.

gpt-image-2-4k

Note: A 4K high-resolution image generation model billed at 0.5-1 yuan per image. It is suitable for high-resolution posters, product images, cover images, finished designs, and commercial visuals with rich detail.

GPT 5.6 Series 🌟🌟🌟🌟

gpt-5.6-sol

Note: A highly capable model in the GPT 5.6 series. It has a 258K context window, supports image understanding, and is suitable for complex Q&A, coding, long-document understanding, and multimodal tasks.

gpt-5.6-terra

Note: A balanced model in the GPT 5.6 series. It has a 258K context window, supports image understanding, and is suitable for everyday Q&A, coding assistance, document processing, and multimodal tasks.

gpt-5.6-luna

Note: An entry-level promotional model in the GPT 5.6 series. It has a 258K context window, supports image understanding, and is suitable for frequent use, basic Q&A, document processing, and image understanding tasks.

xAI / Grok 🌟🌟🌟🌟

grok-4.5

Note: A powerful general-purpose xAI Grok model. It has a 500K context window, supports image understanding, and is suitable for complex Q&A, coding, long-document understanding, and image understanding tasks.

Anthropic (Recommended for Complex Tasks) 🌟🌟🌟🌟🌟

claude-haiku-4-5-20251001

Note: A lightweight, fast Claude model with a 256k context window and image understanding support. It is suitable for summarization, classification, quick Q&A, and low-cost, high-frequency use.

claude-sonnet-5

Note: Claude's next-generation flagship high-capability model. It has a 1M context window, supports image understanding, and is suitable for complex reasoning, coding, multi-turn agents, high-quality writing, and long-document tasks.

claude-fable-5

Note: Claude's top-tier high-capability model. It has a 1M context window, supports image understanding, and is suitable for the most valuable complex reasoning, deep research, long-running agents, long-form writing, and mission-critical engineering tasks.

claude-opus-4-6

Note: A stable, highly capable Opus model. It has a 1M context window, supports image understanding, and is suitable for complex reasoning, serious writing, code architecture, and long-document analysis.

claude-opus-4-7

Note: An enhanced model in the Opus series. It has a 1M context window, supports image understanding, and is particularly suitable for deep coding, complex planning, and analytical tasks that require multiple rounds of refinement.

claude-opus-4-8

Note: An advanced Opus model with a 1M context window and image understanding support. It is suitable for high-value complex tasks, long-running agents, rigorous reports, and high-quality engineering delivery.

China-Based AI Models

China-Based Aggregated Model (Recommended for High Concurrency) 🌟🌟🌟🌟🌟

claude-sonnet-4-6

Note: The AI API Proxy's top recommended high-concurrency model. It has a 1M context window, supports image understanding, and delivers fast response and token generation speeds. It is suitable for agents, workflows, automation scripts, and high-volume requests.

Qwen (Recommended for Long Context and Agents) 🌟🌟🌟🌟🌟

qwen3.6-plus

Note: A cost-effective Qwen model with a long 1M context window and image understanding support. It is suitable for long-document understanding, codebase review, knowledge-base Q&A, agent tool calls, and multimodal office workflows.

qwen3.7-plus

Note: Qwen's next-generation balanced agent model. It has a 1M context window, supports image understanding, and offers comprehensive tool-calling capabilities. It is suitable for OpenClaw, Claude Code, Hermes, large codebases, document processing, and image understanding tasks.

qwen3.7-max

Note: Qwen's flagship reasoning model. It has a 1M context window and is suitable for complex reasoning, deep code planning, architecture design, long-running agents, and high-value analysis. It is currently officially listed as a reasoning and text generation model, without image understanding support.

qwen3.8-max

Note: The latest high-capability Qwen route available on this gateway. It has a 1M context window, supports image understanding, and is suitable for complex reasoning, coding, long-running agents, research, and multimodal tasks. qwen3.8-max is a platform route ID; keep it distinct from the qwen3-max name in Qwen's public documentation.

Meituan / LongCat (Recommended for Long Context and Coding Agents) 🌟🌟🌟🌟

LongCat-2.0

Note: Meituan's long-context agentic coding model. It has a 1M context window and currently supports text input. It is suitable for codebase understanding, complex planning, long-document reasoning, and long-running agent tasks.

Tencent Hunyuan 🌟🌟🌟🌟

hy3

Note: Tencent Hunyuan 3 reasoning and agent model. It has a 256K context window and is suitable for Chinese-language reasoning, coding, long-document understanding, and automation tasks. It is currently integrated with text-model capabilities.

MiniMax (Recommended for Long Context and Agents) 🌟🌟🌟🌟

MiniMax-M3

Note: A multimodal model with a 1M context window and image understanding support. It is suitable for long documents, codebases, multi-turn agents, and tool-calling tasks.

image-01

Note: A text-to-image model billed per image. It is suitable for posters, article illustrations, product images, avatars, creative assets, and bulk visual content generation.

image-01-live

Note: A dynamic visual generation model billed per second. It is suitable for animated images, character animation, real-time visuals, and video-oriented creative scenarios.

StepFun 🌟🌟🌟🌟

step-3.7-flash

Note: A fast vision-language model with a 256k context window and image understanding support. It is suitable for coding, conversations, search augmentation, image understanding, and lightweight agents.

ByteDance / Doubao 🌟🌟🌟🌟

doubao-seed-2.0-code

Note: ByteDance's coding model. It has a 200k context window, supports image understanding, and is suitable for code generation, debugging, engineering rewrites, and TRAE-style development workflows.

doubao-seed-2.0-pro

Note: ByteDance's enhanced general-purpose model. It has a 128k context window, supports image understanding, and is suitable for complex Q&A, writing, analysis, and multimodal tasks.

Xiaomi MiMo (Recommended for Long Context) 🌟🌟🌟🌟🌟

mimo-v2.5-pro

Note: Xiaomi's professional model. It has a 1M context window and is suitable for long text, complex analysis, code planning, and enterprise automation tasks.

mimo-v2.5

Note: A low-cost general-purpose model with a 1M context window and image understanding support. It is suitable for everyday conversations, rewriting, summarization, document processing, and batch tasks.

DeepSeek (Recommended for Reasoning and Coding) 🌟🌟🌟🌟🌟

deepseek-v4-pro

Note: A highly capable DeepSeek MoE model. It has a 1M context window and is suitable for coding, mathematics, complex planning, deep analysis, and advanced reasoning.

deepseek-v4-flash

Note: A faster, lighter DeepSeek model with a 1M context window. It is suitable for high-frequency Q&A, summarization, quick coding assistance, and cost-sensitive reasoning tasks.

Kimi / Moonshot AI 🌟🌟🌟🌟🌟

kimi-k3

Note: Kimi's next-generation long-context multimodal model. It has a 1M context window, supports image understanding, and is suitable for complex coding and agent tasks, ultra-long document analysis, knowledge Q&A, and image understanding.

Zhipu / Z.ai 🌟🌟🌟🌟

glm-5.1

Note: Zhipu's flagship general-purpose model. It has a 256k context window, supports image understanding, and is suitable for writing, analysis, coding, agent engineering, and multimodal scenarios.

glm-5.2

Note: Zhipu's next-generation powerful multimodal model. It has a 1M context window, supports image understanding, and is suitable for complex writing, analysis, coding, agent engineering, and multimodal tasks.

Recommendations

If you are not sure which model to choose, start with your use case. For everyday use, prioritize low-cost, stable models. Move to higher-multiplier models for complex tasks.

  • Everyday default / high concurrency: claude-sonnet-4-6, gpt-5.6-luna, MiniMax-M3
  • Complex reasoning / high-quality output: gpt-5.6-sol, qwen3.8-max, deepseek-v4-pro
  • Top-tier long-running tasks: claude-fable-5, for truly high-value, long-running, complex agent, coding, and research tasks
  • Fast coding iterations: gpt-5.3-codex-spark, LongCat-2.0, qwen3.7-plus
  • Long-context documents: MiniMax-M3, qwen3.7-plus, LongCat-2.0, deepseek-v4-pro, kimi-k3, glm-5.2
  • Image understanding: Choose models whose descriptions explicitly state that they support image understanding
  • Image generation: gpt-image-2, gpt-image-2-4k, image-01, image-01-live