ionos/Qwen3.8-27B-GPTQ-Int4

Update README.md

Restore the MTP head so vLLM can run MTP speculative decoding

Add tool calling and instruction following results

Add multimodal and German evaluation results

Compact the card: evaluation first, drop prose duplicated from the base model card

Drop table scores: they measure output format compliance, not table reading

Add OmniDocBench recovery table vs BF16 base, in percentages

Add OmniDocBench results; recovery table pending BF16 reference

Add README.md

Add LICENSE

Add files using upload-large-folder tool

initial commit