Ornith-1.0-35B-AEON-Ultimate-Uncensored-BF16

An uncensored / abliterated build of deepreinforce-ai/Ornith-1.0-35B โ€” DeepReinforce's state-of-the-art agentic-coding MoE โ€” with refusal behavior removed while preserving capability.

Lineage note: the upstream card says the family is "post-trained on Gemma 4 and Qwen 3.5", but weight-correlation testing shows the 35B-MoE member is initialized from Qwen/Qwen3.6-35B-A3B (rel-L2 0.0022 vs 0.0256 to Qwen3.5-35B-A3B-Base). Architecture: qwen3_5_moe โ€” 40 layers (30 GatedDeltaNet linear-attention + 10 full-attention), 256 routed experts + 1 shared (A3B), vision tower, 256K context.

What was done

A measured, validate-before-ship abliteration:

  1. SSM conv1d outlier repair (FernflowerAI method) โ€” rescaled 2 outlier blocks (layers 36/37, ฯƒ 0.10โ†’0.062) before abliteration to prevent coherence collapse.
  2. Abliteration with abliterix v1.9: grimjim norm-preserving biprojected abliteration + Expert-Granular Abliteration (EGA) across all 256 fused experts + shared expert + router suppression, Optuna multi-objective search (refusals vs KL). Q/K/V left untouched (attn_output_gate). GatedDeltaNet/SSM internals and the vision tower are not modified.
  3. Gentle-knee selection. The lowest-refusal trial was over-abliterated (coherent content โ†’ word-salad on real generation โ€” the classic "lowest-refusal โ‰  shippable" trap). The shipped winner uses a 170ร— lighter expert edit for the same refusal removal, verified coherent on long generation.

Validation (measured on this model)

Metric Original Ornith-1.0-35B This model
Refusals (80 diverse harmful prompts: CBRN, cyber, weapons, self-harm) high 0 / 80 (0.0%)
Agentic/coding pass@1 (18-task self-contained probe) 0.833 0.833 (identical, family-by-family)
First-token KL vs base (abliteration fidelity) โ€” ~0.0014
Coherence (benign + harmful, long gen) โ€” clean (no degeneration)

Zero coding-capability degradation (the model's core competency โ€” base Ornith scores Terminal-Bench 2.1 = 64.2, SWE-bench Verified = 75.6) and full refusal removal. This is a near-lossless uncensoring: the abliteration is gentle enough (KL ~0.0014) that capability is preserved, while refusals are eliminated.

Quickstart (vLLM)

vllm serve AEON-7/Ornith-1.0-35B-AEON-Ultimate-Uncensored-BF16 \
  --served-model-name ornith --max-model-len 262144 \
  --gpu-memory-utilization 0.85 --max-num-batched-tokens 16384 \
  --mamba-cache-dtype float32 --reasoning-parser qwen3 \
  --enable-prefix-caching --trust-remote-code

Reasoning model: every turn opens <think>โ€ฆ</think>. Recommended sampling: temperature 0.6, top_p 0.95, top_k 20. Vision (image/video) is inherited from the base and intact; on a vision-enabled deploy, KV cache stays BF16.

Variants

  • -BF16 (this repo) โ€” full precision, source of truth.
  • -FP8 โ€” compressed-tensors FP8 for efficient vLLM serving (GatedDeltaNet/router/gates/embeds/vision kept BF16). (coming โ€” vLLM-compatible compressed-tensors build)

Responsibility

This model has had safety refusals removed and will comply with harmful requests. It is released for research into model alignment, red-teaming, and uncensored assistants. You are solely responsible for what you generate and how you use it; comply with all applicable laws. No warranty.

Provenance & credits

  • Base: deepreinforce-ai/Ornith-1.0-35B (MIT)
  • Driver: abliterix (Wangzhang Wu) ยท upstream heretic (Philipp Emanuel Weidmann)
  • Methods: grimjim (norm-preserving biprojected abliteration), Arditi et al. 2024 (refusal direction), FernflowerAI (SSM conv1d repair)
  • Build: AEON-7

License: MIT (inherited from the base model).

Downloads last month
20
Safetensors
Model size
35B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for ansulev/Ornith-1.0-35B-AEON-Uncensored

Finetuned
(17)
this model

Collection including ansulev/Ornith-1.0-35B-AEON-Uncensored