AI Potluck
Back to Gap Map Model components / Image, video, 3D & music generation

Hunyuan3D

Tencent
closed / Overall score: 3.0

Hunyuan3D is Tencent's family of image-to-3D models that generate a shape with one diffusion model and paint its texture with a second. Its current release, Hunyuan3D 3.0, is served only through Tencent Cloud and the Hunyuan3D site. The downloadable releases run from Hunyuan3D 2.0 and its mini and multi-view variants to Hunyuan3D 2.1, which adds a physically based material model and releases its training code.

The current release, Hunyuan3D 3.0, is hosted-only, and Hunyuan3D 2.5 was released as a technical report without weights; Hunyuan3D 2.1 remains downloadable under Tencent's regional community license. It scores 1/closed on the current release, and the older open line trails it. A separate availability attribute that would record that gap is tracked in #727.

Openness

1 medium confidence
1.0
weights
closed(Hunyuan3D 3.0 is offered only through the Tencent Cloud Hunyuan 3D API and the Hunyuan3D site
data
closed(nothing is published for Hunyuan3D 3.0)
code
closed(no code is published for Hunyuan3D 3.0)
license
Proprietary(Hunyuan3D 3.0 is available only under Tencent Cloud service terms)
prior-release
distributed(Hunyuan3D 2.1 (and 2.0): downloadable under the Tencent Hunyuan 3D Community License, which excludes the European Union, the United Kingdom and South Korea, with training code published for 2.1)

Hunyuan3D 3.0, the current release, can only be used through Tencent Cloud, with no weights published. The previous downloadable release, Hunyuan3D 2.1, carries a regional community license that excludes the European Union, the United Kingdom and South Korea. The open line trails the vendor's closed release. A separate availability attribute that would record that gap is tracked in #727.

Adoption

3 high confidence
3.0

Adoption is measured as Hugging Face downloads across the six Hunyuan3D checkpoints, led by Hunyuan3D 2.0.

Capability

3 medium confidence
3.0

Hunyuan3D places in the upper half of the image-to-3D arena for rendered quality and leads its separate board for mesh topology. TRELLIS and Meshy rate higher on rendered quality, the arena's main board.

Verified 2026-09-26