HERETIC 1.4.0MTPAPEXMIT Ornith-1.0-35B-Heretic-MTP-APEX English | 📖 中文文档 Heretic Trial 63 BF16 merge · compatible MTP head · APEX GGUFs + mmproj 🐦 About this releaseOrnith-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 DetailsArchitectureQwen3.5 MoE, multimodalParameters35B total, 3B active per token · 256 experts, 8 activeLayers40 transformer layers + 1 MTP layer (785 BF16 tensors)Context262,144 tokensWeightsBF16 safetensors, 17 shards; audited LoRA mergeLicenseMIT ⚡ Heretic Trial 63 of 120Selected parameters: directionindex=21.66; attention oproj range 1.55–2.05; MLP down_proj range 1.64–1.84.Refusal prescreen / full evaluation2 / 30 · 11 / 100Validation KL mean / p95 / max0.0242
Модальности:
Генерация текста Мультимодальность
Области применения:
Логика и рассуждение Диалог / чат
Задача: Генерация текста
Автор: SC117
Теги: qwen3_5_moe, image-text-to-text, reasoning, agentic-coding, heretic, mtp, apex, quantization
Лайков: 3 | Загрузок: 591
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
