Tier 3 Practical open-weight
What the publisher claims: open-source, open-weights
The one-liner: Apache-2.0 image weights that trade capability for speed without adding license conditions.
The six questions
Base and Turbo weights are downloadable from the publisher's repositories without gating.
Apache-2.0 on both checkpoints, an OSI-approved permissive license with no revenue or use threshold.
A six-billion-parameter model built for eight-step, sub-second generation on local hardware.
Public material describes size, speed, and output quality; training data and recipe are not published.
Limited.
Tongyi-MAI publishes weights with no visible development channel.
License
Apache-2.0 OSI-approved — license source ↗
- Standard Apache-2.0 obligations only: retain notices and state changes
What is actually released
- weights (base and Turbo)
- inference code
- model cards
Why it matters
Demonstrates that an OSI-approved weight license can coexist with essentially no training transparency — permissiveness and reproducibility are separate axes.
Sources
- https://huggingface.co/Tongyi-MAI/Z-Image
- https://huggingface.co/Tongyi-MAI/Z-Image-Turbo
- https://pxz.ai/blog/what-is-z-image-model
Derived from the openness dataset behind But Is It Open? — snapshot 2026-11-07 ↗. Scores are qualitative readings of the cited evidence, not verdicts; tier measures openness, not intelligence.