> ## Documentation Index
> Fetch the complete documentation index at: https://docs.xingchaoyiqing.com/llms.txt
> Use this file to discover all available pages before exploring further.

# All AI Models Available Through GlobalAI OPC Gateway

> Compare all video, image, and text AI models available through GlobalAI OPC, including Sora, VEO, Seedance, Vidu, Grok, GPT, Claude, and Gemini.

GlobalAI OPC gives you access to a curated set of best-in-class AI models across three categories: **video generation**, **image generation**, and **text & chat**. All models are available through a single unified API gateway using standard Bearer token authentication, so you can switch between providers without changing your integration structure.

## Video Generation Models

The table below summarizes every video generation model available through GlobalAI OPC, along with supported resolutions, durations, and key capabilities.

| Model Family                 | Model Variants                                                                                                           | Resolution        | Duration          | Notes                                                                    |
| ---------------------------- | ------------------------------------------------------------------------------------------------------------------------ | ----------------- | ----------------- | ------------------------------------------------------------------------ |
| **Sora** (OpenAI)            | `openAiSora2Plus`, `openAiSora2Pro`                                                                                      | 1280×720          | 4, 8, 12s         | Image-to-video supported; max 1 reference image                          |
| **VEO** (Google)             | `veo_3_1_fast`, `veo_3_1_pro`, `veo_3_1_fast_stable`, `veo_3_1_pro_stable`                                               | Up to 4K          | Varies by variant | Reference images and first/last frame control supported                  |
| **Seedance 1.5** (ByteDance) | `seedance_1_5_pro_480p`, `seedance_1_5_pro_720p`, `seedance_1_5_pro_1080p`                                               | 480p – 1080p      | 2 – 12s           | Audio generation supported                                               |
| **Seedance 2.0** (ByteDance) | `seedance_2_0`, `seedance_2_0_fast`, `seedance_2_0_pro`, `seedance_2_0_fast_pro`                                         | Varies by variant | 4 – 15s           | Multi-modal reference input; asset library management required for faces |
| **Vidu**                     | `viduq3-pro`, `viduq3-turbo`                                                                                             | 540p – 1080p      | 1 – 16s           | Flexible duration; first/last frame support                              |
| **Grok** (xAI)               | `grok_video3`, `grok_video3_pro`, `grok_video3_max`, `grok_video3_stable`                                                | 480p – 720p       | 6 – 30s           | Up to 7 reference images for image-to-video                              |
| **StarVideos O3**            | `starvideos_o3`                                                                                                          | Varies            | 3 – 15s           | Video reference input supported                                          |
| **Videos**                   | `videos`, `videos_fast`                                                                                                  | 720p              | 4 – 15s           | Standard and fast-tier variants                                          |
| **SD2 Manxue**               | `sd2_manxue_720p`, `sd2_manxue_1080p`, `sd2_manxue_2k`, `sd2_manxue_4k`, `sd2_manxue_fast_720p`, `sd2_manxue_fast_1080p` | Up to 4K          | 4 – 15s           | Audio reference supported; asset review required for sensitive content   |

## Image Generation Models

GlobalAI OPC supports two image generation model families, both capable of high-resolution output from text prompts and optional reference images.

| Model Family             | Model Variants                                                                                      | Max Resolution | Aspect Ratios             | Notes                                                             |
| ------------------------ | --------------------------------------------------------------------------------------------------- | -------------- | ------------------------- | ----------------------------------------------------------------- |
| **Nano Banana**          | `nano-banana`, `nano-banana-2`, `nano-banana-pro`, `nano-banana-2-stable`, `nano-banana-pro-stable` | Up to 4K       | Multiple ratios supported | Up to 6 reference images for style or subject guidance            |
| **GPT Image 2** (OpenAI) | `gpt-image-2`, `gpt-image-2-r`                                                                      | Up to 4K       | Multiple ratios supported | Quality tiers: `low`, `medium`, `high`; up to 16 reference images |

## Text & Chat Models

Text and chat APIs use a **synchronous** request/response pattern — no polling required. GlobalAI OPC exposes three API protocols to maximize compatibility with existing SDKs and tools.

| API Protocol                                 | Supported Models                                                                                           |
| -------------------------------------------- | ---------------------------------------------------------------------------------------------------------- |
| **OpenAI Chat API** (`/v1/chat/completions`) | GPT-5.4, GPT-5.5, all Claude variants, all Gemini variants                                                 |
| **Claude Messages API**                      | Claude Haiku 4-5 through 4-7, Claude Sonnet 4-5 through 4-7, Claude Opus 4-5 through 4-7, GPT-5.4, GPT-5.5 |
| **Gemini Native API**                        | Gemini 2.5 Flash Lite, Gemini 2.5 Pro, Gemini 3 Flash, Gemini 3.1 Pro                                      |

## Choosing the Right Model

<Accordion title="I need the fastest video generation">
  Use **`videos_fast`**, **`viduq3-turbo`**, or **`seedance_2_0_fast`** for the lowest latency. These fast-tier variants sacrifice some output quality in exchange for significantly shorter generation times — ideal for prototyping or high-throughput pipelines.
</Accordion>

<Accordion title="I need the highest video quality">
  Use **`sd2_manxue_4k`** for 4K output, **`veo_3_1_pro`** for Google's highest-quality tier, or **`seedance_2_0_pro`** for ByteDance's professional tier. These models produce the best results but take longer to generate.
</Accordion>

<Accordion title="I need to generate video from an image">
  All of the following model families support image-to-video (providing one or more reference images to guide the generated video): **Sora**, **VEO**, **Seedance 1.5 & 2.0**, **Vidu**, and **Grok**. Grok supports up to 7 reference images in a single request. For face-based or sensitive imagery, use Seedance 2.0 or SD2 Manxue with the asset management workflow.
</Accordion>

<Accordion title="I need long video duration">
  Use **Grok** (`grok_video3`, up to 30 seconds) for the longest possible output. **Vidu** supports up to 16 seconds with flexible per-second duration control. Most other models cap out at 12–15 seconds.
</Accordion>

<Accordion title="I need to use the OpenAI Python or Node.js SDK">
  Point your SDK's `base_url` at the GlobalAI OPC endpoint and use the **OpenAI Chat API** protocol. This gives you access to GPT, Claude, and Gemini models all through the familiar `openai.chat.completions.create()` interface without changing your existing code structure.
</Accordion>