OculusMindAI/OculusMind-ToolCall-8B-v2
LoRA-style (fewer layers) vs mistralai/Ministral-3-8B-Instruct-2512-BF16. OculusMindAI/OculusMind-ToolCall-8B-v2 matches part of mistralai/Ministral-3-8B-Instruct-2512-BF16 (309 of its 531 tensors; the rest of mistralai/Ministral-3-8B-Instruct-2512-BF16 has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 34 of 34 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/69, MLP 0/102, Attention 136/136, lm_head 0/1