💡 [REQUEST] - Multiuse benchmarks

[FerLuisxd]

起始日期 | Start Date

5/18/2026

实现PR | Implementation PR

相关Issues | Reference Issues

摘要 | Summary

Enhance the benchmarks to include more general use benchmarks like multi use or code

基本示例 | Basic Example

缺陷 | Drawbacks

未解决问题 | Unresolved questions

[tc-mb]

Hi, thanks for the suggestion.

Our current public benchmarks mainly focus on vision / multimodal understanding and related capabilities. Broader “multi-use” or code-centric suites are not on our near-term evaluation roadmap right now.

That said, community contributions are welcome — if you (or others) have a concrete benchmark setup / scripts that fit MiniCPM-V, feel free to open a PR and we can take a look.