unknown model architecture: 'glm5next'

[RubenGarcia]

koboldcpp 1.121

glm5next is used by GLM-5.3-Flash-Q8_0-00001-of-00008.gguf

[RubenGarcia]

I suppose we need to wait for upstream llama to support it.

[henk717]

That is correct, at the time that we made KoboldCpp 1.121 glm5next was still a PR on their side.
Unsloth also has their own engine fork which they are trying to get upstreamed in ggml-org#27754
Until that merges we can not add support for this. I'm not sure whats taking so long upstream.
Once they do merge it we will get the support automatically, if they end up merging something else unsloth will have to make new quants possibly.