--- library_name: transformers license: mit license_link: https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B/blob/main/LICENSE pipeline_tag: text-generation tags: - qwen3_5_moe - reasoning - agentic-coding - heretic - mtp - apex - quantization - gguf - multimodal base_model: - SC117/Ornith-1.0-35B-Heretic-MTP base_model_relation: quantized ---
HERETIC 1.4.0MTPAPEXMIT

Ornith-1.0-35B-Heretic-MTP-APEX

English | πŸ“– δΈ­ζ–‡ζ–‡ζ‘£

Heretic Trial 63 BF16 merge Β· compatible MTP head Β· APEX GGUFs + mmproj

🐦 About this release

Ornith-1.0-35B is an agentic coding model from DeepReinforce AI, based on Qwen3.5 MoE. This experimental derivative merges the selected Heretic 1.4.0 Trial 63 LoRA into BF16 weights.

It also includes a compatible Qwen3.5 MTP head, a BF16 vision projector, and APEX mixed-precision GGUFs. Refusal-modified models can behave differently from the original; evaluate before deployment. License: MIT.

🧠 Model Details
ArchitectureQwen3.5 MoE, multimodal
Parameters35B total, 3B active per token Β· 256 experts, 8 active
Layers40 transformer layers + 1 MTP layer (785 BF16 tensors)
Context262,144 tokens
WeightsBF16 safetensors, 17 shards; audited LoRA merge
LicenseMIT
⚑ Heretic Trial 63 of 120

Selected parameters: direction_index=21.66; attention o_proj range 1.55–2.05; MLP down_proj range 1.64–1.84.

Refusal prescreen / full evaluation2 / 30 Β· 11 / 100
Validation KL mean / p95 / max0.0242 / 0.1477 / 0.1648
Generation healthPassed Β· top-1 disagreement 2.0%
πŸ“Š BenchLocal Results (APEX-I-Compact)
ModeToolCallBugFindHermesAgentMax / Effective
Thinking β†—100988995.0 / 90.0
No Thinking β†—97978592.2 / 76.2

Latest result cards: Thinking has 5 retry-penalty points; No Thinking has 16.

πŸš€ Usage

llama.cpp (vision + text)

hf download SC117/Ornith-1.0-35B-Heretic-MTP-APEX-GGUF --include "*.gguf" --local-dir ./models ./llama-server -m ./models/Ornith-1.0-35B-Heretic-MTP-APEX-I-Compact.gguf --mmproj ./models/mmproj-Ornith-1.0-35B-Heretic-MTP-BF16.gguf -ngl 99 -c 131072

πŸŽ›οΈ Recommended Settings
General / Codingtemperature=0.6, top_p=0.95, top_k=20
πŸ’‘ What is APEX?

These GGUFs use APEX, an MoE-aware mixed-precision quantization method. It allocates precision according to tensor role and layer position; the MTP layer remains Q8_0 and the matching vision projector remains BF16.

πŸ“¦ APEX Quantization Tiers
FileSizeBest For
*-APEX-I-Quality.gguf21.87 GiBHighest quality
*-APEX-I-Balanced.gguf24.33 GiBBest all-rounder
*-APEX-I-Compact.gguf16.14 GiBBest quality/size ratio
*-APEX-I-Mini.gguf13.29 GiBMost compact
## Links - **Original Model**: https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B - **Heretic**: https://github.com/p-e-w/heretic - **APEX Quantization**: https://github.com/mudler/apex-quant - **BenchLocal Results**: https://scorp1o117.github.io/benchlocal-results/ ## Citation ```bibtex @misc{ornith-35b, title = {{Ornith-1.0-35B}: Agentic Coding, Open to All}, url = {https://deep-reinforce.com/ornith_1_0.html}, author = {{DeepReinforce Team}}, year = {2026} } ```