whatmodel.sh / video / Wan 2.2 A14B
Wan 2.2 A14B
↗ share
+ contribute
Wan 2.2 A14B is an open-weights video model from Alibaba, released July 2025. Its published size is 27B parameters (14B active per token). Licence: apache-2.0.
$ how it compares
The video models ranked either side of Wan 2.2 A14B, out of 139 with a score, then more from Alibaba.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 65 |  Seaweed-7B | 50±6 | comparison only | ||
| 66 |  Luma Dream Machine 1.6 | 50±7 | comparison only | ||
| 67 | Wan 2.2 A14B | 50±2 | api $0.080/sec +local | ||
| 68 |  Runway Gen-4 | 50±4 | comparison only | ||
| 69 |  Veo 2 | 50±2 | comparison only | ||
| more from Alibaba | |||||
| 2 | Wan 3.0 Prime | 80±7 | api $0.14/sec | ||
| 4 | Wan 3.0 | 79±2 | api $0.10/sec | ||
| 9 | HappyHorse 1.1 | 76±3 | api $0.14/sec | ||
$ what it's good at
text to video▆▄▃▄▆▅█▃▂▂▂▄ #35 of 68 · 65% of its score · 4 tests
Trails LTX-2 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LMArena video | 1131.6 ✎ | ▂▂▃▆█▆·▆▅▁▂▇ beats 6/41 | 44% |
| Cosmos HUE human evaluation, text-to-video (Cosmos 3 report) | 88.2% ✎ | █▄▄··▄▄▄▄▄·▄ beats 6/9 | 9.0% |
| WorldJen human preference (Bradley-Terry) | 1518.0 ✎ | █···██·█··██ beats 3/5 | 6.4% |
| VideoVerse overall (Gemini 2.5 Pro judge) | 1085.08 / 1608 ✎ | █·▄··▄·····█ beats 3/5 | 5.7% |
image to video▁▂·▃▅█▅▄▅▆▃▄ #46 of 50 * · 29% of its score · 1 test
Trails Wan 2.1 I2V Turbo, HunyuanVideo-I2V, Hailuo 01 (MiniMax video-01 / I2V-01) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena image-to-video | 1170.2 ✎ | ▁▁▁▁▄▇▃▆▃▂▄█ beats 5/40 | 29% |
other tests1 not in a task · 4.3% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| VideoArgus-Bench T2V (rubric-grounded score) | 84.1% ✎ | █▄·▄█▄▄····▄ beats 5/8 | 4.3% |
$ where to run it
Through an API
api · fal.ai · $0.080/sec · fal.ai ↗ · price sourceUSD per video second at 720p on fal.ai. Product: fal Wan 2.2 A14B text-to-video endpoint. API model: fal-ai/wan/v2.2-a14b/text-to-video. Settings: resolution=720p. fal counts video seconds at 16 frames per second (the endpoint's default frame rate). 580p $0.06, 480p $0.04 per video second. Currency: USD. fal.ai model page, observed 2026-09-25.
Other ways to run it
version
- Run it as published · needs ~31.2 GB · from github.com/leejet
sd-cli -M vid_gen --diffusion-model models/Wan2.2-T2V-A14B-LowNoise-Q4_K_M.gguf --high-noise-diffusion-model models/Wan2.2-T2V-A14B-HighNoise-Q4_K_M.gguf --vae models/wan_2.1_vae.safetensors --t5xxl models/umt5-xxl-encoder-Q8_0.gguf -p "a lovely cat" --cfg-scale 3.5 --sampling-method euler --steps 10 --high-noise-cfg-scale 3.5 --high-noise-sampling-method euler --high-noise-steps 8 -v -n "色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走" -W 832 -H 480 --diffusion-fa --offload-to-cpu --video-frames 33 --flow-shift 3.0
Weights and documentation: huggingface.co/Wan-AI ↗
$ how the score was worked out · LMArena leaderboard dataset 74%, Cosmos 3: Omnimodal World Mod… 9%, WorldJen: An End-to-End Multi… 6% and 2 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 LMArena leaderboard dataset | 2 | 74% |
 Cosmos 3: Omnimodal World Models for Physical AI (arXiv:2606.02800v4) | 1 | 9% |
| WorldJen: An End-to-End Multi-Dimensional Benchmark for Generative Video Models | 1 | 6% |
| VideoVerse: Does Your T2V Generator Have World Model Capability to Synthesize Videos? | 1 | 6% |
| VideoArgus: Agentic Rubric-Grounded Unified Evaluation for Video Generation and Editing (arXiv:2608.05485v1) | 1 | 4% |
| prior · pull toward the middle | 1% | |
| weighted average | = 49.8 |
$ published speed measurements · 3
Reference results for the hardware, engine and settings below. They aren't verified speeds of the downloads above or predictions for your machine.
| NVIDIA H100/H800 | bf16 (--convert_model_dtype; shared config) | 1041.5 s/video | Wan reference implementation · T2V-A14B; 720P; 1 GPU; --offload_model True; no prompt extension; --convert_model_dtype; mean after warmup PNG time column manually visually verified. Combined GPU labels preserved; no claim that the run was specifically the H100 rather than H800. CPU model offload is explicit in settings; not misclassified as no offload. | source ↗ |
| NVIDIA H100/H800 | bf16 (--convert_model_dtype; shared config) | 326.9 s/video | Wan reference implementation · T2V-A14B; 480P; 1 GPU; --offload_model True; no prompt extension; --convert_model_dtype; mean after warmup PNG time column manually visually verified. Combined GPU labels preserved; no claim that the run was specifically the H100 rather than H800. CPU model offload is explicit in settings; not misclassified as no offload. | source ↗ |
| NVIDIA A100/A800 | bf16 (--convert_model_dtype; shared config) | 2735.7 s/video | Wan reference implementation · T2V-A14B; 720P; 1 GPU; --offload_model True; no prompt extension; --convert_model_dtype; mean after warmup PNG time column manually visually verified. Combined GPU labels preserved; no claim that the run was specifically the H100 rather than H800. CPU model offload is explicit in settings; not misclassified as no offload. | source ↗ |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json