Framedash
Framedash / Home / API

MiniMax H3 Max API reference

Facts for builders integrating text-to-video. Framedash does not speak for MiniMax or fal.ai — verify everything against live provider docs.

Get credits

Model identifiers

Per the public listings (October 2026), H3 Max is a post-trained variant of MiniMax H3, jointly released by MiniMax and fal.ai on 27 August 2026. Use these ids when you configure requests:

ProviderModel idNotes
MiniMax platformMiniMax-H3-MaxOfficial MiniMax API name
OpenRouterminimax/hailuo-3-maxSilent MP4 on OpenRouter
fal.aiH3 Max endpointsfal marketed ~3 s for a 5 s clip on its stack; list rates match below

Capabilities

Request shape (video)

Typical JSON body fields:

FieldValues
modelProvider model id (see table above)
promptNatural-language scene description
duration5–15 (integer seconds)
resolution480p or 768p
aspect_ratioOne of the six ratios listed above

Submit POST {base}/videos → response includes id, status, and often a polling_url. Poll GET {base}/videos/{id} on the same API base you submitted to (do not follow a different host from polling_url). Statuses include pending, in_progress, completed, failed, cancelled, and expired. On completion, read unsigned_urls[0] with your API key. The poll payload may include usage.cost in USD — that is the billed amount for the clip.

List price per second of output

These are list rates per second of generated video, not Framedash pack prices:

Provider480p768p
MiniMax platform$0.05/s$0.08/s
OpenRouter$0.05/s$0.08/s
fal.ai list$0.05/s$0.08/s

Worked cost examples (list API)

Multiply per-second rate by duration. Examples at list price:

Duration480p total768p total
5 s$0.25$0.40
10 s$0.50$0.80
15 s$0.75$1.20

fal ran a 35% discount promo through 31 October 2026 on its blog. MiniMax H3 base (not Max) lists 768p at $0.08/s and 2K at $0.13/s on the MiniMax platform.

Speed expectations

fal positions H3 Max for speed on its own infrastructure (~3 s for a 5 s clip in its marketing). Through our OpenRouter + Cloudflare AI Gateway path, a 480p 5 s clip took about 16 s end to end in our October 2026 test. Your latency will vary by queue, region, and provider.

Limits and integration notes

Framedash vs raw API

Framedash wraps the same OpenRouter model id with Turnstile verification, daily spend caps, credit wallets, and a file proxy. You still pay list provider rates through our reservation system; packs bundle credits above raw API cost. When generation is off, this page remains a factual reference — not a support channel for MiniMax or fal.

Try the model in our UI: MiniMax H3 Max generator · Framedash credit pricing · H3 vs H3 Max

What people ask before they generate

What is the official MiniMax H3 Max API model id?

On the MiniMax platform the id is MiniMax-H3-Max. On OpenRouter it is minimax/hailuo-3-max. Framedash is unofficial; confirm ids in each provider’s current docs before you ship.

How do you submit a video job?

Typical flow: POST to /videos with model, prompt, duration (5–15 whole seconds), resolution (480p or 768p), and aspect_ratio. The response includes an id and status; poll GET /videos/{id} until status is completed or failed.

Where does the MP4 come from?

Completed jobs expose unsigned_urls that require your API bearer token to download. Framedash proxies files through /api/render/:id/file so browser clients never see provider URLs.

Is audio included?

OpenRouter returns silent video for this model. Plan your own soundtrack in post.

Does H3 Max support 2K?

No. H3 Max tops out at 768p. Base MiniMax H3 supports 768p and 2K at higher list rates on the MiniMax platform.

Is this page official MiniMax documentation?

No. Framedash is independent. Use MiniMax, OpenRouter, or fal.ai for authoritative integration details.