LatitudeGames/Harbinger-24B
Full fine-tune (fewer layers) vs mistralai/Mistral-Small-3.1-24B-Instruct-2503. LatitudeGames/Harbinger-24B matches part of mistralai/Mistral-Small-3.1-24B-Instruct-2503 (363 of its 585 tensors; the rest of mistralai/Mistral-Small-3.1-24B-Instruct-2503 has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 40 of 40 layers.
Tensors changed: Embeddings 1/1, Norms 0/81, MLP 120/120, Attention 160/160, lm_head 1/1
10 derivatives checked
- LatitudeGames/Harbinger-24B-GGUF: Same weights, converted
- mradermacher/Harbinger-24B-i1-GGUF: Quantization
- mradermacher/Harbinger-24B-GGUF: Quantization
- MuXodious/Harbinger-24B-absolute-heresy: Abliterated
- vprilepskii/Harbinger-24B-biprojected-norm-preserving-abliterated: Abliterated
- OddTheGreat/Cogwheel_24b_V.2: Full fine-tune
- Aleteian/TerraIncognita-24B: Full fine-tune
- Vortex5/ChaosRose-24B: Full fine-tune
- Delta-Vector/Austral-24B-Base: LoRA-style
- FlareRebellion/DepravedCartographer-v1.0-24b: Full fine-tune