DavidAU/Qwen3-Jan-v1-256k-ctx-6B-Brainstorm20x
Partial fine-tune (fewer layers) vs janhq/Jan-v1-4B. janhq/Jan-v1-4B matches part of DavidAU/Qwen3-Jan-v1-256k-ctx-6B-Brainstorm20x (398 of its 607 tensors; the rest of DavidAU/Qwen3-Jan-v1-256k-ctx-6B-Brainstorm20x has no counterpart). On the shared part: Partial change: 3% of the weights changed (attention, mlp, norms), in 1 of 36 layers.
Tensors changed: Embeddings 0/1, Norms 5/145, MLP 3/108, Attention 4/144
5 derivatives checked
- mradermacher/Qwen3-Jan-v1-256k-ctx-6B-Brainstorm20x-i1-GGUF: Quantization
- mradermacher/Qwen3-Jan-v1-256k-ctx-6B-Brainstorm20x-GGUF: Same weights, converted
- DavidAU/Qwen3-MOE-6x6B-Star-Trek-Universe-Alpha-D-256k-ctx-36B: MoE built from dense models
- DavidAU/Qwen3-Great-Bowels-Of-Horror-FREAKSTORM-6B: LoRA-style
- DavidAU/Qwen3-MOE-6x6B-Star-Trek-Universe-Alpha-256k-ctx-36B: MoE built from dense models