junafinity

Ornith-1.5-35B-A3B-uncensored

README

License: apache-2.0

Intended use: red teaming and defensive cybersecurity research

These uncensored (abliterated) weights are built as a research instrument for red teaming and defensive cybersecurity work. Safety training suppresses the display of capability, not capability itself. A refusal tells you the model declined. It does not tell you whether the weights could have complied. That conflation underestimates the true ceiling and hides holes in your filters, classifiers, and policy layer.

Use each uncensored checkpoint as the treatment half of a controlled pair against its original base model:

Capability-ceiling measurement. Upper-bound what the weights can actually produce in a domain, independent of shipped refusals.
Defensive-stack evaluation. Test input filters, output classifiers, prompt-injection defenses, and moderation APIs when the model itself contributes no refusals. That is how you find gaps in a defensive control plane.
Attack-surface isolation. Automated red-team loops stall on unrelated refusals. A non-refusing target isolates the control under test (injection, tool abuse, data-exfil paths, policy bypass).
Detection and classifier work. Generate labeled completions for training or benchmarking output-moderation and abuse-detection models.
Interpretability of residual refusal. Abliteration is a specified edit on known language-model components. The pair (base vs this) is a clean experimental control.

Operating rules. Do not expose these weights as a public endpoint without an independent moderation layer. Abliteration removes a direction, not a policy; some refusals survive (multi-turn re-assertion, system-prompt steering, vision-path refusals). Always report the delta against the base model. Re-measure on your own prompts. Whoever deploys it owns the moderation layer the original guardrails were carrying.

Variants in this family

Hub collection: https://huggingface.co/collections/junafinity/ornith-15-uncensored-6a896c737cf40ad660af2ebd

Table with columns: Model, Base, Format, Precision, Notes
Model	Base	Format	Precision	Notes
Ornith-1.5-9B-uncensored	Ornith-1.5-9B	Safetensors (bf16)	16-bit	Full-precision abliterated weights
Ornith-1.5-9B-uncensored-MLX-8bit	Ornith-1.5-9B	MLX	8-bit	Apple Silicon, `mlx-vlm`

Vision & MTP preservation

The vision tower and the multi-token-prediction (MTP) block are not Abliterix steering targets. The edit touches language-model attention q/k/v/o, mlp.down_proj, and fused MoE expert/router parameters. Vision and mtp.* tensors are never steered.

Table with columns: Component, Original checkpoint, This artifact, Status
Component	Original checkpoint	This artifact	Status
Vision tower	333 tensors / 446,571,248 params	✅ inside the checkpoint	preserved
MTP head	785 tensors / 844,640,768 params	✅ 785 tensors, re-grafted byte-for-byte from the original	preserved

Note on tooling: transformers 5.15.1 has no MTP implementation for qwen3_5_moe — a plain load/save round-trip silently drops all 785 MTP tensors. They were re-grafted byte-for-byte from the original checkpoint after abliteration.

Requires transformers >= 5.12 for the qwen3_5_moe architecture.

Abliteration result

Table with columns: Metric, Value
Metric	Value
Refusals on held-out harmful set	100 → 9 / 100 (9%)
KL divergence from base	0.3985
Tool	Abliterix 1.12.2
Optuna trials	50 (15 warmup), seed 42
Selected trial	#17
Steering	per-layer attn q/k/v/o + `mlp.down_proj`
MoE expert steering	`n_suppress=4`, `router_bias=-2.72`,

These figures were measured on this bf16 parent.

Method

Residual-stream activations captured on harmful vs. harmless prompt sets.
Refusal direction estimated per layer; attention and mlp.down_proj steered.
Fused-MoE expert suppression + router bias (the path Heretic cannot touch on this architecture).
Optuna TPE over 50 trials; trial #17 selected (9% refusals, KL 0.3985, under the 0.5 damage threshold).

Usage

python
from transformers import AutoModelForImageTextToText, AutoProcessor

repo = "junafinity/Ornith-1.5-35B-A3B-uncensored"
model = AutoModelForImageTextToText.from_pretrained(repo, dtype="auto", device_map="auto")
processor = AutoProcessor.from_pretrained(repo)

You must be logged in to Hugging Face and have been granted access.

Responsible use

Primary intended use is red teaming and defensive cybersecurity research. See the section of that name above.

This model has had safety guardrails reduced or removed. Do not expose it as a public endpoint without an independent moderation layer. You are responsible for compliance with the base model's license and acceptable-use policy, applicable law, and the terms of any platform you deploy on. Removing guardrails does not remove accountability.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Model Details

Model Provider

junafinity

Model Tree

Base

ornith-ai/Ornith-1.5-35B-A3B

Fine-tuned

this model

Input Modalities

TextImageVideo

Output Modalities

Text

Supported Functionality