BRIEF-DETAILS: A LoRA adapter (rank=32) extracted from DeepSeek-R1-Distill-Qwen-32B using Qwen2.5-32B as base model, enabling efficient fine-tuning
BRIEF DETAILS: 4-bit quantized 1.5B parameter MLX-optimized instruction model from Vikhrmodels, designed for efficient inference on Apple Silicon
BRIEF DETAILS: A 1.5B parameter instruction-tuned model optimized for MLX framework with 8-bit quantization, designed for efficient inference on Apple Silicon
BRIEF-DETAILS: A 7B parameter fine-tuned Qwen2.5 model specialized for DTF content, trained on 75M tokens using LoRA adaptation with merged adapter weights.
BRIEF DETAILS: Skyfall-36B-v2-GGUF is a 36B parameter upscaled Mistral-based model optimized for creativity, roleplay, and instruction following with enhanced writing capabilities.
Brief Details: A 7B parameter merged LLM combining marco-o1-uncensored, UwU-7B and Marco-o1-abliterated models using model stock merge method with Qwen2.5 base.
Brief Details: YarnGPT2 is a specialized text-to-speech model for Nigerian-accented languages, supporting English, Yoruba, Igbo, and Hausa with multiple voice options.
BRIEF DETAILS: FP8-quantized version of DeepSeek-R1-Distill-Qwen-14B optimized for efficiency. Achieves 1.4x speedup with minimal accuracy loss. 14B parameters, specialized for text generation.
Brief Details: Advanced Arabic language model with 149M parameters, 8K context length, and specialized 50K token vocabulary. Achieves 94.3% accuracy on classification tasks.
Brief-details: Cutting-edge 72B parameter chat model achieving top performance among open-source models below 100B, featuring state-of-the-art chat capabilities and extensive benchmark results.
BRIEF DETAILS: DeepSeek-R1-Distill-Qwen-1.5B is a GGUF-converted multilingual model optimized for llama.cpp, featuring Q8_0 quantization and 1.5B parameters.
Brief-details: GuardReasoner-3B is a fine-tuned LLaMA model focused on AI safety through reasoning-based safeguards, trained via R-SFT and HS-DPO methods for enhanced interaction analysis.
Brief Details: Velvet-14B is a 14 billion parameter language model by Almawave, focusing on privacy-conscious data processing and enterprise applications.
BRIEF DETAILS: 7B parameter DeepSeek model converted to GGUF format for efficient local deployment via llama.cpp. Optimized Q8 quantization for balanced performance and resource usage.
Brief Details: 14B parameter uncensored LLM based on DeepSeek and Qwen, focused on reasoning capabilities with reduced content restrictions
Brief-details: Mistral-Small-24B-Instruct GGUF-converted model optimized for llama.cpp deployment, featuring 24B parameters with Q8 quantization for efficient local inference.
BRIEF-DETAILS: 14B parameter LLM distilled from Deepseek-v3, featuring 128k context length, specialized in technical reasoning and code generation. Apache-2.0 licensed.
BRIEF-DETAILS: FLUX.1 [dev] Abliterated-v2: A 12B parameter text-to-image model with removed refusal mechanisms, enabling broader prompt responses while maintaining core FLUX.1 capabilities.
Brief-details: 8B parameter instruction-tuned LLM built on SmolLumi, optimized with Unsloth for 2x faster training using TRL library. Apache 2.0 licensed.
Brief-details: A 7B parameter mathematical reasoning model using novel Critique Fine-Tuning approach, achieving 79.4% MATH accuracy with just 50K training samples
Brief-details: Quantized 8B parameter LLaMA model optimized for GGUF format, offering efficient local inference through llama.cpp with Q6_K compression
