superwhisper/s1-mini
LoRA-style vs Qwen/Qwen3-0.6B. 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 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/113, MLP 84/84, Attention 112/112
7 derivatives checked
- superwhisper/s1-mini-GGUF: Same weights, converted
- mlx-community/S1-mini-MLX-4bit: Quantization
- mlx-community/S1-mini-MLX-bf16: Same weights, repacked
- mradermacher/s1-mini-GGUF: Same weights, converted
- mlx-community/S1-mini-MLX-8bit: Quantization
- stanthegorilla/S1-mini-Q8_0-Q6_K-GGUF: Quantization
- iniyanai/s1-mini-6bit-mlx: Quantization
3 other repos hold byte-for-byte identical weight files.