QUASAR-QAT/gemma-4-E4B-it-QUASAR-Q4_0-GGUF

Add standard Q4_0 (llama-quantize, no imatrix) baseline row

Add standard Q4_0 (llama-quantize, no imatrix) baseline row

Chart: shorten the Unsloth repack label (same fix as 12B)

Card: opener/twin sentences per review; drop duplicate 4-bit tag

Card: Unsloth UD-Q4_K_XL comparator labeled accurately (verified Q4_0 tensors; file_type=2)

Model card: reviewer pass (2x-KL headline, one-line broader comparison, Downstream benchmarks w/ EVAL.md pointer, compact native-QAT, tightened Run it); metadata: +quantized/4-bit/4-bit precision/weight-only/multimodal/gemma-4-e4b(+-it)/capability tags, license_link; legacy model-index removed

Add citation block

Link every compared checkpoint (Google/Unsloth/LM Studio files, BF16 base, twins) to its exact HF file

QUASAR native Q4_0 GGUF + BF16 mmproj (private draft)

card v1: card-only chart / no chart

QUASAR native Q4_0 GGUF + BF16 mmproj (private draft)

QUASAR native Q4_0 GGUF + BF16 mmproj (private draft)

QUASAR native Q4_0 GGUF + BF16 mmproj (private draft)

QUASAR native Q4_0 GGUF + BF16 mmproj (private draft)

QUASAR native Q4_0 GGUF + BF16 mmproj (private draft)

initial commit