vanta-research/apollo-astralis-2
LoRA-style vs mistralai/Ministral-3-8B-Reasoning-2512. 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 0/119, MLP 0/174, Attention 136/232, other 0/4, lm_head 0/1
2 derivatives checked
- mradermacher/apollo-astralis-2-i1-GGUF: Quantization (fewer layers)
- mradermacher/apollo-astralis-2-GGUF: Same weights, converted (fewer layers)