Model Card on HF 🤗: https://huggingface.co/Undi95/Phi4-abliterated
Phi4 Abliterated Benchmarks
MMLU Pro: 47.57
GPQA: 10.74
MUSR: 8.93
BBH: 54.12
IFEval: 66.18 vs 88 (so35)-24.8%
MATH Lvl 5: 37.01
LLME Score: 0.36227
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
What is the LLM Explorer Rank (Score)
Phi4 Abliterated Parameters and Internals
| LLM Name | Phi4 Abliterated |
| Repository 🤗 | https://huggingface.co/Undi95/Phi4-abliterated |
| Model Size | 14.7b |
| Required VRAM | 29.4 GB |
| Updated | 2025-03-12 |
| Maintainer | Undi95 |
| Model Type | phi3 |
| Model Files | |
| Model Architecture | Phi3ForCausalLM |
| Context Length | 16384 |
| Model Max Length | 16384 |
| Transformers Version | 4.47.1 |
| Tokenizer Class | GPT2Tokenizer |
| Padding Token | <|endoftext|> |
| Vocabulary Size | 100352 |
| Torch Data Type | bfloat16 |
Best Alternatives to Phi4 Abliterated
Best Alternatives |
Context / RAM |
Downloads |
Likes |
|---|---|---|---|
| Phi 4 | 16K / 29.4 GB | 525498 | 1891 |
| Microsoft Phi 4 | 16K / 29.4 GB | 605 | 55 |
| Flashlight V1.1 | 16K / 29.4 GB | 30 | 0 |
| Phi 4 RP V0 | 16K / 29.5 GB | 48 | 3 |
| Phi 4 AbliteratedRP | 16K / 29.5 GB | 58 | 3 |
| Phi 4 Abliterated | 16K / 29.4 GB | 130 | 28 |
| DNA R1 | 16K / 29.4 GB | 691 | 30 |
| Phi 4 Open R1 Distill EZOv1 | 16K / 29.4 GB | 218 | 18 |
| Phi 4 Bf16 | 16K / 29.3 GB | 186 | 1 |
| EZO Phi 4 V2.900 | 16K / 29.4 GB | 56 | 0 |
Note: green Score (e.g. "73.2") means that the model is better than Undi95/Phi4-abliterated .
Rank the Phi4 Abliterated Capabilities
🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟
What open-source LLMs or SLMs are you in search of? 44887 in total.
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241227