kmouratidis/Devstral-Small-2507-Rebased-Vision

Same weights, converted (fewer layers) vs mistralai/Devstral-Small-2507. mistralai/Devstral-Small-2507 matches part of kmouratidis/Devstral-Small-2507-Rebased-Vision (363 of its 585 tensors; the rest of kmouratidis/Devstral-Small-2507-Rebased-Vision has no counterpart). On the shared part: Same weights stored in another dtype (BF16 → F16), plus 222 new tensors.

Tensors changed: Embeddings 0/1, Norms 0/81, MLP 0/120, Attention 0/160, lm_head 0/1

2 derivatives checked

On the Hub · Downloads on Model Pulse