mradermacher/Step3-VL-10B-Base-GGUF

Same weights, converted (fewer layers) vs stepfun-ai/Step3-VL-10B-Base. mradermacher/Step3-VL-10B-Base-GGUF matches part of stepfun-ai/Step3-VL-10B-Base (399 of its 1066 tensors; the rest of stepfun-ai/Step3-VL-10B-Base has no counterpart). On the shared part: Same weights stored in another dtype (BF16 → F16/F32).

Tensors changed: Norms 0/145, MLP 0/108, Attention 0/144, Embeddings 0/1, lm_head 0/1

On the Hub · Downloads on Model Pulse