Does MiroThinker-H1 use the same agent framework as MiroThinker 1.7? | MiroThinker-H1 是否使用与 MiroThinker 1.7 相同的 agent 框架？

[panademo]

Congrats to the MiroMind team on the MiroThinker 1.7 and H1 launch, and thank you so much for open-sourcing such impressive models and sharing your research!

A quick question about the proprietary MiroThinker-H1: is H1 essentially the same agentic framework as MiroThinker 1.7 with a different base model, or does it also use a materially different agent framework/system design in addition to a different base model?

I’m mainly trying to understand whether the performance gap is driven primarily by the base model or by changes to the agentic setup as well.

Thanks again for your amazing work! 🙏

---

恭喜 MiroMind 团队发布 MiroThinker 1.7 和 H1！也非常感谢你们持续开源如此出色的模型并分享相关研究成果！

我有一个关于专有模型 MiroThinker-H1 的小问题：H1 是否本质上使用与 MiroThinker 1.7 相同的 agent 框架，只是底层基础模型不同？还是说除了基础模型不同之外，它在 agent 框架 / 系统设计上也有比较大的不同？

我主要想了解，目前性能差距主要来自底层模型本身，还是也来自 agent 架构方面的改进。

再次感谢你们出色的工作！🙏