Poor name. Its hard to find information about it due to it being confused with h3 vision model
For the image generation they should of called it something besides h3.
This is simply a derivative model based on the H3 video model. The next release will be the H3.1 video model, and we may then see an H3.1 image model as well. The purpose is to make it easier to maintain consistent assets across the entire video production process.
This is simply a derivative model based on the H3 video model. The next release will be the H3.1 video model, and we may then see an H3.1 image model as well. The purpose is to make it easier to maintain consistent assets across the entire video production process.
H3.1 video model when? soon or it takes sometime?
This is simply a derivative model based on the H3 video model. The next release will be the H3.1 video model, and we may then see an H3.1 image model as well. The purpose is to make it easier to maintain consistent assets across the entire video production process.
H3.1 video model when? soon or it takes sometime?
Approximately 1 to 2 months.
August 26, 2026: During an earnings call, MiniMax’s CEO mentioned that H3.1 and M3.1 were nearing completion, while M3.1 Flash, which was part of the same planned release batch, had already been officially launched.
In its official article published on July 31, 2026, titled “MiniMax H3: Breaking the Boundaries of Tasks and Modalities,” MiniMax explicitly outlined three key directions for future R&D:
- Integrating the Capabilities of the M-Series Models
Official statement:
We will work toward integrating the capabilities of the M-Series models into the next version of the H-Series.
This is an important strategic direction, indicating that the next-generation H-Series will feature enhanced multimodal understanding capabilities.
- Increasing Model Scale (Scaling)
Official statement:
Due to limitations in model size, there is still room for improvement in H3's performance across certain capabilities. Scaling is a clear direction for us.
This indicates that MiniMax considers the current scale of H3 to be a limitation and views increasing model size as a clear technical direction.
- Improving Visual Fidelity and Resolution
Official statement:
The visual fidelity of H3 still needs improvement in certain scenarios. We will continue striving for higher resolution and finer image quality.
In addition to the above, I believe future updates will likely include longer one-shot generation durations of over 30 seconds and higher limits on the number of reference inputs.