Burnt-Toast/ms3.2-24b-longform
LoRA-style (fewer layers) vs mistralai/Mistral-Small-3.2-24B-Instruct-2506. Burnt-Toast/ms3.2-24b-longform matches part of mistralai/Mistral-Small-3.2-24B-Instruct-2506 (363 of its 585 tensors; the rest of mistralai/Mistral-Small-3.2-24B-Instruct-2506 has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 40 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 120/120, Attention 160/160, lm_head 0/1
3 derivatives checked
- mradermacher/ms3.2-24b-longform-i1-GGUF: Quantization
- mradermacher/ms3.2-24b-longform-GGUF: Quantization
- Vortex5/MS3.2-24B-Penumbra-Aether: Full fine-tune