Z.aiVideo model
CogVideoX1.5-5B T2V
Clip lengthRank 33 of 63
10slongest single cliphigher is better
Summary
CogVideoX1.5-5B T2V is a video model from Z.ai. CogVideoX1.5-5B T2V is a text-to-video checkpoint; the separate I2V variant remains its own catalog entry.
Head-to-head comparisons
vs.CogVideoX1.5-5B T2V vs. FLUX.3 VideoNot enough data for a result.CogVideoX1.5-5B T2VFLUX.3 VideoImage to videon/a1,449Text to videon/a1,4935 s clipn/a$0.85vs.CogVideoX1.5-5B T2V vs. Dreamina Seedance 2.5 720pNot enough data for a result.CogVideoX1.5-5B T2VDreamina Seedance 2.5 720pImage to videon/a1,477Text to videon/a1,4745 s clipn/an/avs.CogVideoX1.5-5B T2V vs. Dreamina Seedance 2.0 720pNot enough data for a result.CogVideoX1.5-5B T2VDreamina Seedance 2.0 720pImage to videon/a1,475Text to videon/a1,4795 s clipn/a$0.76vs.CogVideoX1.5-5B T2V vs. MiniMax H3Not enough data for a result.CogVideoX1.5-5B T2VMiniMax H3Image to videon/a1,495Text to videon/a1,4605 s clipn/a$0.65vs.CogVideoX1.5-5B T2V vs. Veo 3.1Not enough data for a result.CogVideoX1.5-5B T2VVeo 3.1Image to videon/a1,398Text to videon/a1,3645 s clipn/a$2.00vs.CogVideoX1.5-5B T2V vs. Veo 3.1 FastNot enough data for a result.CogVideoX1.5-5B T2VVeo 3.1 FastImage to videon/a1,385Text to videon/a1,3625 s clipn/a$0.50
Show all comparisons (94)
All specifications
Architecture
Input and output
Licence and access
- Model license
- The weights use the separate CogVideoX LICENSE. Apache 2.0 for the code is not the model license.
- Source: Model license
Further details
- Checkpoint
- Canonical current Hugging Face ID: zai-org/CogVideoX1.5-5B. The former THUDM URL redirects to the current zai-org namespace.
- Source: Checkpoint
- Input mode
- Text-to-video checkpoint; the separate I2V variant is not merged into this record.
- Source: Input mode
- Generation limits
- The official model matrix lists 1360 × 768, 5 or 10 seconds, and 16 frames per second.
- Source: Generation limits
Frequently asked questions about CogVideoX1.5-5B T2V
How long can a video with CogVideoX1.5-5B T2V be?
Up to 10 seconds in one go. Resolution: 1360 × 768.
According to Z.ai
- Longest clip
- 10s
- Resolution
- 1360 × 768
- Open weights
- yes
Sources and data date
| Type | Evidence and data date |
|---|---|
| Additional source | Z.ai (retrieved November 1, 2026) · Z.ai · Editorial description reviewed November 1, 2026 |
| Additional source | Z.ai (retrieved November 1, 2026)12 |