Best text-to-video models for a 32GB graphics card
A leaderboard of the best text-to-video models for a 32GB graphics card (with 32GB of system RAM). The best one that runs on it right now is Cosmos3-Nano, followed by HunyuanVideo 1.5 and Wan 2.2 A14B. Models are ranked on human votes and video generation tests. Models that run on it show the download that fits and the command to run it.
Scores use reviewed text to video evaluations only. Compare scores within this task. Models without task measurements are excluded.
+ more
$ terminal
↗ share
detecting your OS…
Score vs API price
● open weights ◆ API only ─┘ best score for the price · x: USD per second of video (log)
| # | model | task score↓ | where↓ | speed↓ | sources |
|---|
| # | model | task score↓ | where↓ | speed↓ | sources |
|---|---|---|---|---|---|
| 1 |  Gemini Omni 1.1 Flash | 82±2* | api $0.10/sec |  | |
| 2 | Gemini Omni Flash | 81±2* | api $0.10/sec |  | |
| 3 |  FLUX 3 Video | 80±2* | api $0.17/sec | | |
| 4 |  Dreamina Seedance 2.0 | 78±2* | api $0.30/sec | | |
| 5 |  Wan 3.0 | 78±2* | api $0.10/sec | | |
| 6 |  Dreamina Seedance 2.5 | 78±2* | api $0.30/sec | | |
| 8 |  MiniMax H3 | 76±2* | api $0.13/sec +local | | |
| 9 | HappyHorse 1.1 | 75±7* | api $0.14/sec | | |
| 10 |  HappyHorse 1.0 | 74±2* | api $0.14/sec | | |
| 12 |  Veo 3.1 | 69±2 | api $0.40/sec | | |
| 13 | Veo 3.1 Fast | 69±2 | api $0.10/sec | | |
| 15 |  Grok Imagine Video | 68±2* | api $0.050/sec | | |
| 17 | Wan 2.7 | 67±2* | api $0.10/sec | | |
| 19 | Wan 2.6 | 66±2 | api $0.10/sec | | |
| 20 | Seedance 1.0 Pro Fast | 63±10* | api $0.049/sec | | |
| 21 | Seedance 1.5 Pro | 61±2* | api $0.052/sec | | |
| 22 | Wan 2.5 Preview | 60±2* | api $0.10/sec | | |
| 23 |  PixVerse V5.6 | 59±2* | api $0.045/sec | | |
| 24 |  Runway Gen-4.5 | 58±2* | api $0.12/sec | | |
| 25 |  Kling 2.5 Turbo Pro | 58±2* | api $0.070/sec | | |
| 26 | Kling 2.6 Pro | 58±2* | api $0.070/sec | | |
| 29 |  P-Video | 57±2* | api $0.020/sec | | |
| 30 | Hailuo 2.3 | 57±2* | api $0.047/sec | | |
| 31 | Kling O1 Pro | 57±2* | api $0.11/sec | | |
| 32 | Seedance 1.0 Pro | 56±2* | api $0.052/sec | | |
| 33 | Hailuo 02 | 55±2* | api $0.047/sec | | |
| 34 |  Kandinsky 5.0 Pro | 54±2* | api $0.040/sec | | |
| 35 |  Cosmos3-Nano · bf16 | 54±4* | local (slow) | | |
| 36 |  HunyuanVideo 1.5 · bf16 | 54±2 | local | | |
| 38 | PixVerse 5.5 | 53±9* | api $0.060/sec | | |
| 39 | Kling 2.1 Master | 53±2* | api $0.28/sec | | |
| 40 |  LongCat-Video | 53±7* | api $0.040/sec | | |
| 42 | Wan 2.2 A14B · gguf-q4_k_m ×2 experts + umt5 q8_0 (sd.cpp) | 52±2 | local | | |
| 44 | Wan 2.1 14B · gguf-q8_0 + umt5 q8_0 (sd.cpp) | 50±2 | local |  | |
| 46 | Kandinsky 5.0 T2V Lite · bf16 | 49±2* | local | | |
| 47 |  LTX-2 19B Distilled | 49±7* | api $0.020/sec | | |
| 48 | Cosmos-1.0-Diffusion-7B-Text2World · bf16 | 48±3 | local | | |
| 50 | Hailuo 01 (MiniMax video-01 / I2V-01) | 47±5* | api $0.083/sec | | |
| 52 | Wan 2.2 TI2V 5B · fp16 + umt5 q8_0 (sd.cpp) | 47±4 | local | | |
| 53 |  CogVideoX-5B · bf16 | 47±2 | local |  | |
| 57 |  Pyramid Flow (SD3) · bf16 | 45±3* | local | | |
| 58 |  T2V-Turbo-v2 · fp32 | 45±7*† | local | | |
| 61 | Wan 2.1 T2V 1.3B · fp16 + umt5 q8_0 (sd.cpp) | 43±4 | local | | |
| 62 | LTX-2.3 | 43±7* | api $0.040/sec +local | | |
| 63 |  Mochi 1 · bf16 | 42±2 | local | | |
| 65 | HunyuanVideo | 41±3 | api price unknown +local | | |
| 66 |  Pika 2.2 | 41±2 | api $0.040/sec | | |
| 68 | T2V-Turbo (VC2) · fp32 | 41±3† | local | | |
| 70 | CogVideoX-2B · fp16 | 39±2 | local | | |
| 71 | CogVideoX1.5-5B · bf16 + sequential CPU offload | 38±3 | local (slow) | | |
| 72 |  Allegro · bf16 | 38±2 | local | | |
| 73 |  VideoCrafter2 · fp32 | 37±2 | local | | |
| 76 |  LaVie (base T2V) · fp16 | 35±2 | local | | |
| 77 | LTX-Video 0.9 (2B) · bf16 | 35±3* | local | | |
| 79 |  Open-Sora 1.1 · bf16 | 34±6* | local | | |
| 81 |  NOVA-0.6B T2V · fp16 | 32±3* | local | | |
| 82 |  Hotshot-XL · fp16 | 31±2 | local | | |
| 85 | ZeroScope v2 576w · fp16 | 30±3 | local | | |
| 87 | VideoCrafter1 T2V · official script (checkpoint as stored) | 29±2 | local | | |
| 89 | Open-Sora 1.2 · bf16 | 28±3* | local (slow) | | |
| 90 |  ModelScope T2V 1.7B · fp16 | 28±2 | local | | |
| 91 |  LVDM (text-to-video) · fp32 | 28±3* | local | | |
| 93 | ZeroScope v2 XL (576w → XL pipeline) · fp16 + model CPU offload (576w base, then XL vid2vid) | 27±2 | local | | |
| 94 | CogVideo 9B (original) · fp16 (--fp16), both stages | 26±7* | local (slow) | | |
| — | Kandinsky 6.0 Video Lite · bf16 + model CPU offload | not yet calibrated | local (slow) | | |
| — | LTX-2.5 · fp8 | not yet calibrated | local | |
+ 31 more don't fit your machine or budget · show them
97 of 97 catalogue models listed here · sources and licences
* new or thinly tested: ranked on too few benchmark results · † only the maker’s own numbers so far · ~RAM is estimated; full context may need more memory