Qwen 3.8 MAX for Sisyphus ? Time has come?

[aresstokrat]

Mb now he is gonna make it?
Especially the time when we now have local Qwen 3.8 27B which beats Opus 4.6 Max in several benchmarks.

[CytITDchen]

Don't think so, Sisyphus is not just depend on benchmark but depend on model behavior, check out agent-model-matching guide

[aresstokrat]

I did before i wrote it here.
Since they migrated from Claude (banned) to support from ChatGPT.
I think QWEN is also kind of anti marketing right now.

[CytITDchen]

OK, maybe it could be a good idea, I wish the OMO official teams or someone could test this soon!