Prototype · closed beta
unOpen.AI
In closed beta: Roxy, a private AI assistant with permanent memory running on infrastructure the user owns. Brand, site, product, backend and conversational design.
Model families
Video models.
Ranked by people, weekly.
Every text-to-video and image-to-video model in the Artificial Analysis arenas, where people pick between two anonymous clips made from the same prompt. Refreshed every week; each model has its own page.
All Model Families →
LEADING MODELS
The highest-rated video models.
Ranked by blind preference votes in the Artificial Analysis text to video arena. Each model has its own page with every arena it is ranked in.
#1 · HeyGen
HeyGen Video 1
- Text to video
- 1143
- Image to video
- 1104
- Released
- Oct 2026
#2 · Fal
MiniMax H3 Max
- Text to video
- 1140
- Image to video
- 1125
- Released
- Aug 2026
#3 · Fal
FLUX 3 Max
- Text to video
- 1135
- Image to video
- 1058
- Released
- Oct 2026
#4 · Alibaba
Wan 3.0
- Text to video
- 1133
- Image to video
- 1085
- Released
- Aug 2026
TOP TEN
The top ten, side by side.
Elo rating from blind preference votes · Higher is preferred more often · Linear scale from 0 to 1,300
- HeyGen Video 1
- 1,143
- MiniMax H3 Max
- 1,140
- FLUX 3 Max
- 1,135
- Wan 3.0
- 1,133
- Utopai X
- 1,130
- MiniMax H3 (768p)
- 1,120
- Bach 1.0 Pro
- 1,109
- P-Video-2-Pro
- 1,103
- Dreamina Seedance 2.5
- 1,101
- FLUX 3
- 1,092
Elo is relative: a 100-point gap means voters preferred the higher model about 64% of the time. Gaps inside the confidence intervals in the table below are not meaningful, and preference votes say nothing about price, licence or speed.
Source: Artificial Analysis arenas. Checked 30 October 2026. Measured independently by Artificial Analysis, not by QQuantum.ai or the providers.
EVERY RANKED MODEL
All 37 video models, ranked.
Sort by any column. Select a model for its full page.
| 1 | HeyGen Video 1 | HeyGen | Oct 2026 | 1143 -16/16 | 1104 -17/17 | 5,570 |
|---|---|---|---|---|---|---|
| 2 | MiniMax H3 Max | Fal | Aug 2026 | 1140 -19/19 | 1125 -19/19 | 2,934 |
| 3 | FLUX 3 Max | Fal | Oct 2026 | 1135 -19/19 | 1058 -19/19 | 2,998 |
| 4 | Wan 3.0 | Alibaba | Aug 2026 | 1133 -17/17 | 1085 -19/19 | 4,515 |
| 5 | Utopai X | Utopai Studios | Sep 2026 | 1130 -18/18 | — | 2,967 |
| 6 | MiniMax H3 (768p) | MiniMax | Jul 2026 | 1120 -17/17 | 1111 -19/19 | 4,411 |
| 7 | Bach 1.0 Pro | Video Rebirth | Sep 2026 | 1109 -18/18 | 1120 -19/19 | 2,993 |
| 8 | P-Video-2-Pro | Pruna AI | Sep 2026 | 1103 -18/18 | 1103 -19/19 | 2,977 |
| 9 | Dreamina Seedance 2.5 | ByteDance Seed | Jul 2026 | 1101 -17/17 | 1129 -19/19 | 4,602 |
| 10 | FLUX 3 | Black Forest Labs | Jul 2026 | 1092 -17/17 | 1041 -19/19 | 3,737 |
| 11 | Dreamina Seedance 2.0 | ByteDance Seed | Mar 2026 | 1085 -17/17 | 1094 -20/20 | 3,720 |
| 12 | Gemini Omni Flash 1.1 | Aug 2026 | 1068 -17/17 | 1024 -19/19 | 4,701 | |
| 13 | Agnes-Video-2.5 | Sapiens AI | Aug 2026 | 1057 -17/17 | 1115 -19/19 | 3,651 |
| 14 | Grok Imagine Video 1.5 | SpaceXAI | May 2026 | 1036 -17/17 | 1054 -19/19 | 3,737 |
| 15 | HappyHorse-1.1 | Alibaba-ATH | Jun 2026 | 1033 -16/16 | 1023 -18/18 | 4,692 |
| 16 | HappyHorse-1.0 | Alibaba-ATH | Apr 2026 | 1032 -17/17 | 1011 -19/19 | 3,637 |
| 17 | Wan2.7-260612 | Alibaba | Jun 2026 | 1022 -16/16 | — | 4,648 |
| 18 | MAGI-2 Preview | Sand.ai | Sep 2026 | 1005 -16/16 | — | 5,958 |
| 19 | Kling 3.0 1080p (Pro) | KlingAI | Mar 2026 | 1000 0/0 | 1000 0/0 | 4,475 |
| 20 | PixVerse V6 | PixVerse | Mar 2026 | 990 -17/17 | 969 -19/19 | 3,677 |
| 21 | Kling 3.0 Omni 1080p (Pro) | KlingAI | Mar 2026 | 984 -17/17 | 1021 -19/19 | 3,631 |
| 22 | Grok Imagine Video 1.5 Lite | SpaceXAI | Aug 2026 | 978 -17/17 | 998 -19/19 | 4,150 |
| 23 | Veo 3.1 | Jan 2026 | 960 -17/17 | 955 -19/19 | 3,692 | |
| 24 | Veo 3.1 Fast | Jan 2026 | 959 -17/17 | 934 -19/19 | 4,403 | |
| 25 | Veo 3.1 Lite | Mar 2026 | 943 -17/17 | 923 -19/19 | 3,668 | |
| 26 | LTX-2.5 Fast | Lightricks | Aug 2026 | 929 -17/17 | 918 -19/19 | 3,627 |
| 27 | LTX-2.5 Pro | Lightricks | Aug 2026 | 927 -17/17 | 908 -19/19 | 3,669 |
| 28 | Runway Gen-4.5 (720p) | Runway | Dec 2025 | 909 -17/17 | 901 -19/19 | 3,563 |
| 29 | Vidu Q3 Pro | Vidu | Jan 2026 | 901 -17/17 | 913 -19/19 | 4,083 |
| 30 | Vidu Q3 Turbo | Vidu | Mar 2026 | 899 -17/17 | 926 -19/19 | 4,074 |
| 31 | Kling O1 Pro (January) | KlingAI | Jan 2026 | 890 -18/18 | 867 -20/20 | 3,471 |
| 32 | LTX-2.3 Fast | Lightricks | Mar 2026 | 845 -18/18 | 822 -20/20 | 3,780 |
| 33 | LTX-2.3 Pro | Lightricks | Mar 2026 | 835 -18/18 | 814 -20/20 | 3,875 |
| 34 | Pika 2.5 | Pika Art | Nov 2025 | 835 -18/18 | 851 -20/20 | 3,438 |
| 35 | Boreal | Creatify | Sep 2026 | 821 -18/18 | — | 3,565 |
| 36 | P-Video | Pruna AI | Mar 2026 | 802 -19/19 | — | 3,596 |
| — | Hyperflow H3 | Video Rebirth | Sep 2026 | — | 1116 -20/20 | — |
Elo from the Artificial Analysis arenas: people choose between two anonymous outputs of the same prompt, and the ratings follow from those votes. The figure after each score is its 95% confidence interval; gaps inside it are not meaningful. Rank and votes are for text to video. Refreshed weekly; not measured by QQuantum.ai. Select a heading to sort.
How we evaluate
Video models differ on what a short arena clip cannot show: how long a shot holds together, whether a character stays the same across cuts, how well a still image is animated without distortion, and what each second costs to render. We shortlist from this table and test on storyboards from the actual project.
Check the provider's terms for commercial rights, likeness and voice rules, and content provenance labelling before publishing. Generation time and per-second pricing vary widely, so budget a test run before committing a schedule to one model.
Questions
Questions about Video Generation Models.
What do the two ratings mean?
Text to video rates clips generated from a written prompt; image to video rates clips that animate a supplied still. Both come from blind side-by-side votes in the Artificial Analysis arenas.
Which video model should we use?
Shortlist from the table, then test on your own storyboard: shot length, consistency across shots, motion quality and cost per second. Arena clips are short, so longer work needs its own test.
Can we use generated video commercially?
That depends on the provider's terms for your access route. Check output rights, likeness rules and provenance labelling before publishing.
