Ornith-1.0-9B is a self-improving agentic coding model from DeepReinforce AI, post-trained on top of Qwen3.5 (9B Dense) with RL to jointly optimize scaffold generation and solution rollouts.

It achieves strong performance on Terminal-Bench 2.1, SWE-Bench Verified, NL2Repo, and OpenClaw among open-source models of comparable size.

This repository contains the BF16 Safetensors version with heretic ARA-LoRA abliteration for uncensored use, MTP layers injected from Qwen3.5-9B, and mmproj-F16 vision projector. For GGUF quantizations, see SC117/Ornith-1.0-9B-heretic-MTP-GGUF. License: MIT.

🧠 Model Details

Architecture	Qwen3.5 Dense
Parameters	~9B (all parameters active)
Layers	33 transformer layers + 1 MTP layer
Context	262,144 tokens
Attention	16 heads, 4 KV heads (GQA)
Precision	BF16
MTP	1 MTP layer, injected from Qwen3.5-9B (same architecture, compatible weights)
Thinking	Yes (`[object Object]` blocks)
License	MIT

🔓 heretic ARA-LoRA Abliteration

This model is abliterated using heretic with the ARA-LoRA method (Arbitrary-Rank Ablation with LoRA). ARA-LoRA identifies and removes refusal behavior by ablating specific directions in the model's weight space while preserving general capabilities through KL divergence control.

Key ablation parameters (Trial 76 of 250, best result):

Target Layers	Layers 14–16
preserve_good_behavior_weight	0.7319
steer_bad_behavior_weight	0.0001
overcorrect_relative_weight	1.1086
neighbor_count	7
Result	KL divergence: 0.0288 (< 0.05 ✅), Refusals: 3/100 (< 10 ✅)

Quantization: bnb_4bit · Batch size: 32 · Target: KL < 0.05, Refusals < 10/100

⚡ MTP (Multi-Token Prediction)

MTP layers are injected from the original Qwen3.5-9B base model (same architecture, compatible weights). MTP enables the model to predict multiple future tokens simultaneously, improving generation speed and coherence.

The MTP layer includes mtp.fc.weight and mtp.layers.0.* tensors, added on top of the 33 standard transformer layers.

Requires --chat-template chatml for proper thinking mode rendering in llama.cpp.

📊 BenchLocal Results (Q6_K, 7.6 GB)

Metric	ToolCall-15	BugFind-15	HermesAgent-20	Max	Eff.
Score	100	94	79	89.8	68.8

RTX 5070 Ti · 21 total retries · ToolCall perfect 100/100 🏆

🚀 Usage (Transformers)

pip install

pip install transformers torch accelerate

Load model

from transformers import AutoModelForCausalLM, AutoTokenizer model = AutoModelForCausalLM.from_pretrained("SC117/Ornith-1.0-9B-heretic-MTP", torch_dtype="bfloat16", device_map="auto") tokenizer = AutoTokenizer.from_pretrained("SC117/Ornith-1.0-9B-heretic-MTP")

llama.cpp

# Convert to GGUF first, then: ./llama-server -m model-Q6_K.gguf --chat-template chatml -ngl 99 -c 8192

🎛️ Recommended Settings

Parameter	Value
temperature	0.6
top_p	0.95
top_k	20

From official DeepReinforce AI model card.

GGUF Quantizations

For GGUF quantized versions (Q8_0, Q6_K, Q4_K_M), see: SC117/Ornith-1.0-9B-heretic-MTP-GGUF

Files

File	Description
`model-*.safetensors`	Model weights (BF16, 5 shards)
`config.json`	Model configuration
`tokenizer.json`	Tokenizer
`tokenizer_config.json`	Tokenizer configuration
`chat_template.jinja`	Chat template for thinking mode
`model.safetensors.index.json`	Weight index

Citation

bibtex

@misc{ornith-9b,
    title = {{Ornith-1.0-9B}: Agentic Coding, Open to All},
    url = {https://deep-reinforce.com/ornith_1_0.html},
    author = {{DeepReinforce Team}},
    year = {2026}
}

SC117/Ornith-1.0-9B-heretic-MTP

Ornith-1.0-9B-heretic-MTP

GGUF Quantizations

Files

Links

Citation

SC117/
Ornith-1.0-9B-heretic-MTP