LTX-2.5 vs MiniMax H3: license lines and run requirements
LTX-2.5 vs MiniMax H3 from vendor pages: 10M vs 20M USD revenue lines, excluded territories, Python and CUDA needs, frame rules, audio, and what Sume lists.

The sharpest difference between LTX-2.5 and MiniMax H3 is the license, not the specs. LTX-2.5 is free under the LTX-2.x Community License below 10 million dollars in revenue and needs a paid agreement above that. MiniMax H3 requires written authorization above 20 million US dollars in yearly revenue, and its license defines the applicable territory as worldwide except the US, EU, UK and Republic of Korea.
How do they compare on the pages that matter?
| Item | LTX-2.5 | MiniMax H3 |
|---|---|---|
| Revenue line | 10M: paid Commercial Use Agreement above it | 20M USD: written authorization above it |
| Territory | not mentioned in the card text I read | US, EU, UK, South Korea excluded |
| Software | Python 3.12 or higher, CUDA 12.7 or higher | Linux, Python 3.12 for the SGLang reference |
| Frame rule | num_frames % 8 == 1; size divisible by 32 | 17n + 5 frames at 24 FPS |
| Audio | synchronized audio with video | 32 kHz stereo |
| Length on the card | not stated in the text I read | 4 to 15 seconds |
What does each ask you to display?
H3's license section III.3.a requires a notice reading "Powered by MiniMax H3" on products or services developed with it. I did not find an equivalent line on the LTX-2.5 card, but the card points to a license file, so read it before launch.
What should a small team do first?
Check the license against your own situation before you check the speed. A team in the US, the EU, the UK or South Korea is in the license's excluded territories, so check with MiniMax before using the H3 weights. A team with revenue near 10 million dollars should read the LTX terms about the paid agreement before it builds on the weights.
If the license is fine, run the same five prompts on each and compare time, memory and look. Where a license blocks one of them, you do not need that comparison at all. None of this is legal advice; ask counsel if the stakes are high.
Which ones can you call by id on Sume?
minimax-h3 and minimax-h3-max are in the catalog. LTX is not (catalog code, read 2026-10-09). The hosted minimax-h3 row takes 5 to 15 seconds at 480p or 768p, and a hosted call is a different thing from running the open weights. For a full read of what a company outside the excluded territories can do, see the earlier license comparison, and the video docs for the hosted fields.
Sources
Related posts
More in Comparisons
- Luma Ray 3.2 alternative: one /v1/generations route vs Sume's two
Luma's API sends images and video through POST /v1/generations. Sume splits them into /v1/videos and /v1/images and does not list ray-3.2 in its docs.
- Luma Ray3.2 allows 16 keyframes; Sume offers first and last frame
Luma lists up to 16 keyframes per clip on its Ray3.2 API. Sume does not list Luma; its video models take a first and a last frame through frame_images.
- Luma Ray 3.2 reframe vs LTX reframe: 60-second 1080p, $21.60 vs $12.00
Reframing a 60-second clip to 1080p costs $21.60 on Luma's page (36 cents a second) and $12.00 on LTX's (20 cents). Lower tiers and Sume's $0.02 crop too.
- MAI-Transcribe-2 at $0.54 vs Sume STT at $0.60 an hour: the 6-cent gap
MAI-Transcribe-2-Streaming is $0.54 an audio hour through 2026; Sume STT is $0.60 an hour. At 1,000 hours the gap is $60. They suit different jobs.
Written by Sume