HunyuanVideo 1.5
Video Available Full comparison ↗ tencent/hunyuan-video-1-5 · by Tencent
· Diffusion transformer (DiT)
Pricing — 1 offering(s)
Standard
- $0.375 / clip
Historical
2026-06-13 → 2026-09-19
Correction
Corrected 2026-09-26: Modelglass first recorded a wrong price for these dates; this row replaces it, and the…
Corrected 2026-09-26: Modelglass first recorded a wrong price for these dates; this row replaces it, and the wrong row is no longer served.
- $0.375 / clip Current 2026-09-20 → present
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how HunyuanVideo 1.5 fits into a cost-aware routing setup
See how →Capability profile
How HunyuanVideo 1.5 rates across core capability dimensions, with the task-level evidence behind each rating.
Full-attention 3D DiT produces excellent frame coherence; widely noted as one of the strongest open-weight models on VBench temporal metrics.
Strong motion realism; competitive with commercial closed models.
Dual-stream text conditioning improves following of complex prompts.
No audio.
13B parameter model is slow on consumer hardware; fal.ai hosted inference is practical but expensive per clip at $0.375 (5 seconds, $0.075/s).
Up to ~10 seconds via API.
T2V and I2V only; no editing of existing clips exposed via API or fal.ai.
Ratings are estimated — limited independent data is available for this model.
Benchmarks
| Benchmark | Score | Config | Source |
|---|---|---|---|
| vbench | 85.9 | Approximate; verify against current leaderboard. HunyuanVideo was top-ranked among open-weight models at the time of its release. | leaderboard ↗ |
Operator guidance
Best open-weight option for temporal consistency and motion quality. Self-hosting is practical on A100/H100 hardware. Via fal.ai it costs $0.375 per 5-second clip ($0.075/s) — choose based on which provider's terms suit. Not suitable for latency-sensitive or high-volume workflows due to generation speed.
Use cases
- Premium-quality self-hosted video generation (open-weight)
- Research requiring strong temporal consistency
- Hosted production use via fal.ai where $0.375 per 5-second clip is acceptable
Provenance notice
- Tencent-origin model; subject to Chinese data regulations
- Open-weight release enables self-hosting but terms still apply
Limitations
- Slow inference; requires substantial GPU (13B params)
- fal.ai price is $0.375 per 5-second clip ($0.075/s), higher than budget open-weight options such as Wan 2.5 on fal ($0.05/s)
- No audio generation
- Tencent-origin model; content policy may differ from Western providers