Brief Details: MobileLLM-1B is a 1.01B parameter efficient language model optimized for on-device use, featuring GQA and shared embeddings with 2k context.
Brief-details: WestLake-7B-v2 is a 7B parameter LLM specializing in role-play and text generation, achieving 74.68% average score on key benchmarks with strong performance in HellaSwag (88.65%) and Winogrande (86.98%).
Brief Details: A 7B parameter coding-focused LLM pre-trained on 2T tokens with 4K context window, fine-tuned on 2B instruction tokens for code generation.
Brief-details: A specialized anime-style Stable Diffusion model focused on creating round-eyed female characters with low-height proportions and detailed eye highlights
Brief Details: 15B parameter SQL generation model that outperforms GPT-3.5-turbo, converting natural language to SQL with 77.5% accuracy.
Brief-details: A powerful 72B parameter LLM based on Qwen2, fine-tuned for conversations, coding, and function calling. Features uncensored responses and 128k context window.
Brief-details: Meta's latest 70B parameter LLM, instruction-tuned for superior performance. Features GQA attention, 15T token training, and enhanced code capabilities. Matches/exceeds GPT-3.5.
Brief Details: A powerful 22B parameter dense language model derived from MOE compression, featuring 32k context length, uncensored capabilities, and strong performance in math, coding, and multi-turn conversations.
Brief-details: Redmond-Puffin-13B is a commercially available Llama-2 based model fine-tuned on 3K high-quality examples, achieving SOTA performance on GPT4ALL benchmarks.
Brief-details: A 340B-parameter reward model by NVIDIA for evaluating AI responses across 5 dimensions: helpfulness, correctness, coherence, complexity, and verbosity.
BRIEF DETAILS: A specialized fine-tuned transformer model designed to improve LoRA training performance on Flux-Dev, trained on 3M high-quality images across two epochs.
Brief-details: Pygmalion-13B is a conversational LLaMA fine-tune focused on dialogue generation, requiring XOR decoding with original LLaMA weights for deployment.
Brief-details: A powerful 9.4B parameter language model specialized in long-form content generation, capable of producing 10,000+ word texts in English and Chinese
Brief-details: XEUS is a powerful multilingual speech encoder covering 4000+ languages, using E-Branchformer architecture with 577M parameters, designed for universal speech recognition.
Brief-details: A powerful 60.8B parameter MoE model combining two 34B models, achieving strong performance on various benchmarks with multi-lingual capabilities and efficient architecture.
Brief Details: MistralTrix-v1: An 8.99B parameter model fine-tuned with DPO, achieving top performance among 7B LLMs. Features FP16 precision and English language support.
Brief Details: A Chinese BERT-based sentence embedding model fine-tuned on NLI data, optimized for semantic similarity tasks with 13.4K+ downloads and Apache 2.0 license.
Brief-details: Specialized Japanese-to-Chinese translation model optimized for light novels and visual novels, built on Qwen/Baichuan architecture with extensive ACGN-domain training.
Brief-details: Language model with 67B parameters trained on 2T tokens, supporting English/Chinese text generation. Features Grouped-Query Attention and commercial use license.
Brief Details: A specialized ControlNet model for Stable Diffusion that enables brightness control and image colorization, with recommended weights of 0.4-0.9 for optimal results.
BRIEF DETAILS: Russian-language Llama-3 based chatbot (8B params) with specialized prompt format and strong performance in language tasks. Built for conversational AI.
