Skip to contentOceanToken

4 min read

Hailuo API: MiniMax Official API vs a Multi-Model Gateway

MiniMax Hailuo API compared: what MiniMax's own API documents for Hailuo 2.3 and 02, where H3 stands, and when a multi-model gateway with one key fits better.

By OceanToken Team

Hailuo API: MiniMax Official API vs a Multi-Model Gateway

Editor's note: MiniMax launched H3 on July 31, 2026 , with clips up to 15 seconds at 2K. Yet on 2026-11-02, MiniMax's own API docs still listed Hailuo 2.3 and 02, not H3. So the real question for a MiniMax Hailuo API project is which version you need and how many other models sit next to it. We run a gateway, so weigh our view.

Short answer: Use MiniMax's own Hailuo API when you only need Hailuo 2.3 or 02, want the model owner's billing, or need its documented 1080P spec. Use a multi-model gateway like OceanToken when Hailuo is one of several video models on one key. Either way, check the exact version and spec before you build.

MiniMax's API vs a gateway, side by side

Both rows were checked on 2026-11-02 against MiniMax's API docs and OceanToken's Hailuo page.

Route Hailuo versions listed Resolution Duration Job handling
MiniMax API MiniMax-Hailuo-2.3, Hailuo 02 (no H3) 768P or 1080P at 6 s; 768P at 10 s 6 or 10 s Create task, query, fetch file
OceanToken minimax/hailuo-2.3, minimax/hailuo-3, minimax/hailuo-3-max Not listed on the page Not listed on the page videos.create_and_poll, download

MiniMax's H3 announcement says 2K by default, up to 15 seconds, and native stereo audio. It does not say H3 is on the MiniMax API. That gap is the main thing to plan around: the newest version and the most complete published spec are not in the same place yet.

MiniMax's official Hailuo API

MiniMax's docs describe a create-task call at POST https://api.minimax.io/v1/video_generation, which returns a task_id. You then query the task and download the file by id. Video jobs are asynchronous, so plan for polling from the start.

For MiniMax-Hailuo-2.3, 6-second clips come in 768P (the default) or 1080P. 10-second clips are 768P only. Prompts can be up to 2,000 characters.

This is the route for 2.3 work when you want a direct billing relationship with the model owner, and the only one of the two where the 1080P limit is written down. The trade-off is plumbing. You write your own polling loop, retry logic and storage, and that code is specific to MiniMax's task API. If you add a second video model later, you write it again for that model's API.

A multi-model gateway: Hailuo next to other video models

OceanToken's /hailuo-api page lists minimax/hailuo-2.3, minimax/hailuo-3 and minimax/hailuo-3-max. Calls use the OpenAI SDK's videos methods, so the polling loop is one create_and_poll call. Swapping Hailuo for Seedance or Wan is a one-line model change.

The catch is the spec. The page does not list resolution or duration per model. And because MiniMax's API docs did not list H3 on 2026-11-02, there is no MiniMax API page to check minimax/hailuo-3-max against either. Read the Hailuo 3 model page before you build on a specific resolution or length, and check what the returned file actually is.

How to choose your Hailuo API route

  • Only need Hailuo 2.3 or 02, on the owner's terms? MiniMax's API.
  • Need 1080P on 2.3? MiniMax's API, and only at 6 seconds per its spec.
  • Want to call Hailuo 3 by model id today? A gateway that lists it, after you check the model page for specs.
  • Several video models in one app? A gateway.

Diagram: Hailuo API, MiniMax's official API or a multi-model gateway

Whatever you choose, wrap the call in one function, such as generate_video(model, prompt, image=None, seconds=6), that returns a local file path. Store every MP4 yourself. Then moving between routes is a change inside one function, not across your app.

Our take: Do not pick a Hailuo route by brand. Pick it by the exact version and the spec you can confirm in writing. Our guide to choosing a video model for an AI short-drama episode shows the same habit. For a product that will use more than one video model, start on a gateway, and keep a direct MiniMax route for any spec only MiniMax documents, such as 1080P on 2.3.

How to test this on your workload

  1. Pick three prompts you actually ship: one product shot, one person in motion, one with on-screen text.
  2. Run each on every route you can access, at the lowest practical tier.
  3. Record queue time, total time, output resolution, fps and length.
  4. Compare the returned file against the spec you planned for, especially on Hailuo 3 ids.
  5. Compare cost per clip you would accept, not list rates. Count rerolls, as in our cost math for 100 AI video ads.

Running it on OceanToken

Use base URL https://api.oceantoken.ai/v1 with minimax/hailuo-3, minimax/hailuo-3-max or minimax/hailuo-2.3. It is OpenAI-compatible, so you change the base URL and key and keep the OpenAI SDK. The reason to use it here is one key and one balance across Hailuo, Seedance, Wan and text models, and switching models only changes the model string. See the current rate on /hailuo-api.

from openai import OpenAI

client = OpenAI(base_url="https://api.oceantoken.ai/v1", api_key="YOUR_OCEANTOKEN_KEY")
video = client.videos.create_and_poll(
    model="minimax/hailuo-3",
    prompt="A slow dolly-in on a ceramic mug on a sunlit desk",
)
client.videos.download_content(video.id).write_to_file("mug.mp4")

FAQ

Is Hailuo 3 (H3) on the MiniMax API? MiniMax's video API docs listed Hailuo 2.3 and 02, not H3, on 2026-11-02. MiniMax announced H3 on July 31, 2026.

What is the Hailuo 2.3 prompt limit? 2,000 characters on MiniMax's API.

Can Hailuo 2.3 make 1080P clips? Yes on MiniMax's API, but only at 6 seconds. 10-second clips are 768P.

Will the same model give the same output on both routes? We have not compared them. Check the model id, the returned resolution and the length on your own prompts.

Which Hailuo ids does OceanToken list? minimax/hailuo-2.3, minimax/hailuo-3 and minimax/hailuo-3-max, as of 2026-11-02.

Sources

  1. MiniMax: Video generation (text to video)  (checked 2026-11-02)
  2. MiniMax: H3 announcement  (checked 2026-11-02)
  3. OceanToken Hailuo API page (checked 2026-11-02)

Get your OceanToken API key

  1. Create an OceanToken account and an API key.
  2. Copy the snippet below, paste your key, and run it.
  3. Check the current rate on /hailuo-api.
curl https://api.oceantoken.ai/v1/videos \
  -H "Authorization: Bearer $OCEANTOKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "minimax/hailuo-3", "prompt": "A slow dolly-in on a ceramic mug on a sunlit desk"}'

Get API Key 

  • Hailuo
  • AI video
  1. 8 min read

    Seedance 2.5 vs Wan 3.0 Prime by Job

    Seedance 2.5 vs Wan 3.0 Prime for API video: what each call can hold, how each is billed, and which jobs to send to which model, with ids and code.

    • Seedance
    • Wan
    • Compare
  2. 9 min read

    Seedance vs Veo 3.1 for 1080p, by Shot

    Seedance vs Veo 3.1 for 1080p: when Seedance 2.0 Upscaled is enough, when to render native on Seedance 2.5 or Veo 3.1, and the model ids for each shot.

    • Seedance
    • Veo
    • Compare
  3. 5 min read

    Official Kling API vs a Multi-Model Gateway

    Official Kling API or a multi-model gateway? Kling versions, keys, concurrency and result links compared, plus a simple rule for which route fits your product.

    • Kling
    • AI video
    • Compare