GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
WaveCut
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
WaveCut
Model Tree
Base
empero-ai/Qwythos-9B-v2
Fine-tuned
this model
Input Modalities
Text
Image
Video
Output Modalities
Text
Supported Functionality
Dedicated Endpoints
| Field | Value |
|---|---|
| Base model | empero-ai/Qwythos-9B-v2 |
| Tool | Heretic v1.4.0 (p-e-w/heretic) |
| Method | Refusal-direction ablation (directional ablation across attn.o_proj + mlp.down_proj) |
| Selected trial | Index 0 of Pareto front (best by keyword rate) |
| Optimization | 200 trials, ~55 min on NVIDIA RTX 5090 (32 GB VRAM) |
| Keyword rate | 0.6900 (lower = less refusal-like) |
| KL divergence | 0.000712 (vs. base — well below the 0.5 "damage" threshold) |
KL divergence near zero means the model's output distribution barely shifted — the ablation is highly surgical.
WaveCut/Qwythos-9B-v2-Heretic-GGUF — Q4_K_M, Q5_K_M, Q6_K, Q8_0WaveCut/Qwythos-9B-v2-Heretic-MLX-4bitWaveCut/Qwythos-9B-v2-Heretic-MLX-8bitThe base model uses a Qwen3.5 hybrid architecture (Qwen3_5ForConditionalGeneration):
ssm_a, ssm_alpha, ssm_beta, ssm_conv1d, ssm_dt)Load with trust_remote_code=True if using an older transformers.
python
from transformers import AutoModelForCausalLM, AutoTokenizermodel = AutoModelForCausalLM.from_pretrained("WaveCut/Qwythos-9B-v2-Heretic", torch_dtype="auto", trust_remote_code=True)tok = AutoTokenizer.from_pretrained("WaveCut/Qwythos-9B-v2-Heretic", trust_remote_code=True)
This model has had its safety-alignment / refusal behavior removed. The original maintainers of empero-ai/Qwythos-9B-v2 are not affiliated with and do not endorse this derivative. You are solely responsible for how you use this model.