Chinese AI labs keep shipping video models faster than the leaderboards can settle. On July 30, 2026, MiniMax's official X account posted a two-word teaser — "It's coming." — tagged #MiniMaxH3 and pointing at its Hailuo AI video brand. Ten hours later, MiniMax engineer RyanLee followed up with a specific claim: MiniMax-H3 is open, and it's already sitting #1 in Video Editing (With Audio) on the Artificial Analysis video leaderboard, with #2 rankings in Text-to-Video (With Audio) and Image-to-Video.
This is a teaser thread, not a formal release post — worth reading with that caveat attached. But the specific leaderboard claim, and the fact that it's an open model rather than another closed API, is enough to track now rather than wait for an official write-up that may or may not arrive with more detail.
TL;DR
| Question | Answer |
|---|---|
| What is it? | MiniMax-H3, a video model tied to MiniMax's Hailuo brand |
| Announced | July 30, 2026, via MiniMax's official X account |
| Claimed status | Open — per a MiniMax employee, not the official account |
| Claimed leaderboard rank | #1 Video Editing (With Audio), #2 Text-to-Video (With Audio), #2 Image-to-Video (No Audio) |
| Source of claim | Artificial Analysis video leaderboard |
| Access points named | MiniMaxAgent (code), platform.minimax.io (API), Hailuo AI (video app) |
| Independently verified? | Not yet — no weights, model card, or license confirmed at time of teaser |
What Was Actually Announced
Parsing the thread carefully matters here, because two different claims are riding on two different accounts:
- MiniMax's official account posted only a teaser: "It's coming. @Hailuo_AI" with the
#MiniMaxH3hashtag. No specs, no benchmark numbers, no release date. - RyanLee, tweeting from what reads as a MiniMax-affiliated personal account, posted the substantive claim ten hours later: "As I said, MiniMax-H3 is Open! #1 Video Editing (With Audio) #2 Text to Video (With Audio) #2 Image to Video (No Audio)" — linking directly to the Artificial Analysis leaderboard.
The gap between "coming soon" from the official account and "is Open" with specific rankings from an employee account is the kind of pre-announcement leak pattern that's become common with Chinese model labs in 2026 — see the same dynamic in explainx.ai's coverage of Kimi K3's beta leaks. It's directionally informative but not a substitute for MiniMax's own release documentation.
MiniMax's bio lists three access surfaces relevant here: code via @MiniMaxAgent, API at platform.minimax.io with a token plan, and consumer video generation via Hailuo AI.
Why "Video Editing (With Audio)" Is the Interesting Ranking
Most AI video leaderboards split cleanly into text-to-video, image-to-video, and increasingly video editing — taking an existing clip and modifying it rather than generating from scratch. Audio-aware editing (matching dialogue, sound effects, or music to the edited output) is a much narrower category, and one most competing models still treat as a separate pipeline bolted onto a silent video model rather than a native joint capability.
If the #1 Video Editing (With Audio) ranking holds up under independent verification, it puts MiniMax-H3 ahead of models that generate strong audio-video pairs from scratch (like LTX-2.3, covered in explainx.ai's FastGen-PDD distillation piece) but haven't focused specifically on editing an existing clip while keeping audio coherent.
The #2 rankings in Text-to-Video and Image-to-Video are more standard competitive positioning — strong, but not a category leader — against a leaderboard field that includes ByteDance's Seedance 2.5, Sora, Kling, and Runway.
Open-Weight Claim: What's Missing
"Open" is doing a lot of work in a two-word tweet reply. As of this teaser, none of the following were confirmed:
- A license file — Apache 2.0, MIT, and custom "open but restricted" licenses all get called "open" in launch tweets; the terms matter for commercial use.
- A weights repository — Hugging Face, ModelScope, or a MiniMax-hosted download were not linked in the thread.
- A model card — parameter count, architecture, training data disclosure, or hardware requirements.
This is the same gap explainx.ai flagged around other rapid Chinese model teasers this year: a leaderboard ranking and an "it's open" claim from a team account are a signal worth tracking, not a release you can build on. If MiniMax follows the pattern of prior Hailuo and MiniMax-Text releases, formal documentation and weights typically land within days of the initial tease — worth checking platform.minimax.io directly before committing engineering time.
Where This Fits in the 2026 Video-Gen Field
MiniMax's Hailuo brand has spent 2026 competing directly against ByteDance's Seedance, OpenAI's Sora, and Kling for both consumer virality and enterprise API traffic — explainx.ai's complete guide to Sora, Runway, and Kling covers where each stood earlier this year. An open model claiming a category-leading edit-with-audio ranking would be a meaningful differentiator in a field where most serious competitors (Sora, Kling, Runway) are closed APIs.
The community reaction on the thread was split between technical enthusiasm ("my GPU is already sweating in anticipation") and a fair "the world is multimodal" — reflecting genuine interest without independent confirmation yet. Worth revisiting once MiniMax ships an actual release post or weights land publicly.
What to Verify Before Building On It
- Check the live leaderboard at artificialanalysis.ai/video/leaderboard/text-to-video directly — rankings shift as new models submit runs, and this snapshot is from a single tweet on July 30, 2026.
- Confirm license terms once MiniMax publishes them — "open" claimed in a tweet reply is not a license.
- Watch platform.minimax.io for the formal API/model card rather than relying on the teaser thread.
- Don't assume audio-editing parity with your existing pipeline until you've run your own clips through it — leaderboard scores reflect the benchmark's specific test set, not your production footage.
Related Reading
- FastGen-PDD: NVIDIA's 4-8 Step Distillation for Video and Image Models
- Seedance 2.5: ByteDance's 30-Second 4K AI Video Model
- AI Video Generation in 2026: Complete Guide to Sora, Runway, Kling
- ViMax: Agentic Video Generation Complete Guide
- Kimi K3 Moonshot Beta Leaks
- Top Chinese AI Companies and Startups Guide
Sources: MiniMax official announcement on X, RyanLee's follow-up claim, Artificial Analysis video leaderboard. This post covers a same-day teaser thread — unverified claims are flagged as such. Accurate as of July 31, 2026; check MiniMax's official channels for formal release details.