Menlo/Jan-nano
Full fine-tune vs Qwen/Qwen3-4B. Fine-tune: 90% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 0/1, Norms 98/145, MLP 108/108, Attention 144/144
14 derivatives checked
- unsloth/Jan-nano-GGUF: Same weights, converted
- Menlo/Jan-nano-gguf: Quantization
- bartowski/Menlo_Jan-nano-GGUF: Same weights, converted
- Mungert/Jan-nano-128k-GGUF: Full fine-tune
- Menlo/Jan-nano-128k: Full fine-tune
- mradermacher/Jan-nano-i1-GGUF: Quantization
- mradermacher/Jan-nano-GGUF: Same weights, converted
- DevQuasar/Menlo.Jan-nano-GGUF: Same weights, converted
- prithivMLmods/Jan-nano-F32-GGUF: Same weights, converted
- ReallyFloppyPenguin/Jan-nano-GGUF: Same weights, converted
- yasserrmd/AgenticCoder-4B: Full fine-tune
- yasserrmd/AgentUX-4B: Full fine-tune
- yasserrmd/qwen3-4b-agentic-reasoner: Full fine-tune
- aayanmishra-ml/Athena-R3X-4B: LoRA-style
2 other repos hold byte-for-byte identical weight files.