HunyuanVideo 1.5
A solid pick for self-hosting and full data control.
Overview
HunyuanVideo 1.5 is Tencent's compact text-to-video and image-to-video model, hitting near state-of-the-art open-source quality with just 8.3B parameters — the step-distilled variant renders a clip in 75 seconds on a single RTX 4090. Its Tencent community licence permits commercial use up to 100 million monthly active users, but includes territorial restrictions that may affect EU deployment and should be checked clause by clause. Note that some components (sparse attention, super-resolution modules) haven't been fully released yet.
Skill profile
Not disclosed
Strengths
- Open-source and self-hostable
Limitations
- API pricing not disclosed
Who is it for
- you want to control cost or self-host
- you have GDPR constraints
Ideal use cases
- Rapid video prototyping
- Short marketing clips
- Video-gen R&D
- Low-cost internal demos
- Animating still images
Access & availability
Key specifications
Privacy
Run it locally
| Quantization | Disk | RAM / VRAM | Typical hardware |
|---|---|---|---|
| Q4 · recommended | 4.7 GB | 7 GB | 16GB PC / M1+ Mac / 8GB GPU |
| Q8 · balanced | 8.9 GB | 12 GB | 32GB PC / 24GB Mac / RTX 4070 Ti+ |
| FP16 · max quality | 16.6 GB | 20 GB | 32GB Mac / RTX 3090-4090 |
Estimates for a moderate context. Long contexts need more RAM (KV cache).
Deploy
Copy-ready commands generated from this card. Adjust context length and GPU count to your hardware.
OpenAI-compatible server for production on NVIDIA GPUs.
pip install vllm
vllm serve tencent/HunyuanVideo-1.5 \
--max-model-len 32768 \
--tensor-parallel-size 1 \
--dtype autoAdvanced data · for expertsArchitecture, modalities, detailed cost, full benchmarks▾
| Arena Elo | — |
| MMLU | — |
| GPQA | — |
| HumanEval | — |
| SWE-Bench | — |
| MATH | — |