Brief-details: OpenBioLLM-70B is a state-of-the-art 70B parameter biomedical LLM fine-tuned from Llama-3, achieving 86.06% accuracy across medical benchmarks.
Brief-details: A specialized Stable Diffusion fine-tune that generates robotic artwork, featuring 351 likes and 273 downloads. Uses "nousr robot" prompt prefix.
Brief-details: StableLM-Tuned-Alpha 7B is a fine-tuned decoder-only language model optimized for chat and instruction-following, built on NeoX architecture with 6144 hidden size and 16 layers.
Brief Details: FastChat-T5 (3B params) - Fine-tuned Flan-t5-xl chatbot for commercial/research use. Trained on 70K ShareGPT conversations.
BRIEF DETAILS: A specialized Stable Diffusion 1.5-based model fine-tuned for generating paper cut-style artwork, featuring 360 likes and 551 downloads. Created by Fictiverse.
BRIEF-DETAILS: DepthPro: Apple's advanced monocular depth estimation model delivering high-res metric depth maps in 0.3s, with state-of-the-art boundary accuracy
Brief-details: Uncensored fine-tuned Llama-2 7B model using QLoRA, trained on Wizard-Vicuna dataset. 6.74B params, supports chat-style interactions with less restrictive responses.
Brief-details: Yi-6B is a powerful 6B parameter open-source LLM from 01.ai, trained on 3T tokens with strong bilingual capabilities and Apache 2.0 license
Brief Details: XTTS-v1 is a multilingual voice cloning TTS model supporting 14 languages, requiring only 6-second voice samples for high-quality cross-language voice synthesis.
Brief-details: Genstruct-7B: A 7B parameter instruction-generation model built on Mistral, designed to create high-quality synthetic training data from raw text corpora
Brief Details: Mistral 7B-based multimodal model combining LLaVA 1.5 architecture, trained on 1.2M+ image-text pairs, outperforming Llama 2 13B
Brief Details: Llama-3.1-8B-Omni is a speech-language model enabling seamless speech interaction with LLMs, featuring low-latency response and simultaneous text/speech generation.
Brief Details: Multi-style Stable Diffusion model supporting Archer, Arcane, and Modern Disney styles. Enables style mixing with 573 downloads and CreativeML OpenRAIL-M license.
Brief-details: Pixtral-12b is a multimodal AI model built on Mistral that can process both text and images, featuring advanced vision encoding capabilities with GELU and 2D ROPE.
Brief Details: Advanced 141B parameter MoE model based on Mixtral-8x22B, achieving strong performance in reasoning and multilingual tasks with 52.72% IFEval accuracy.
BRIEF-DETAILS: A comprehensive LORA model series focused on creating realistic Asian faces across different ethnicities, featuring Korean, Japanese, Taiwanese, Chinese and Thai variants.
BRIEF-DETAILS: Text-to-image diffusion model optimized for high-quality image generation with simple prompts. Created by Predogl and piEsposito with commercial usage rights.
Brief-details: Classic hand-drawn cartoon style LoRA for Stable Diffusion XL, specialized in creating whimsical tiny characters with 390 likes and 2.8K+ downloads.
Brief Details: A 13B parameter cybersecurity-focused LLM based on LLaMA-2, designed for offensive and defensive security applications with exceptional reasoning capabilities
Brief-details: BioMedLM (2.7B params) is a specialized biomedical language model trained on PubMed data, achieving SOTA 50.3% accuracy on MedQA tasks.
Brief-details: ViLT model fine-tuned for visual question answering on VQAv2 dataset, offering efficient vision-language processing without convolution or region supervision
