Brief Details: REBEL: 406M parameter BART-based model for relation extraction, achieving SOTA on NYT (93.4 F1) and CoNLL04 (76.65 F1) benchmarks
BRIEF DETAILS: StableLM Base Alpha 7B is a powerful 7B-parameter language model with 4096 sequence length, trained on 1.5T tokens for enhanced context understanding and generation capabilities.
Brief Details: NuExtract: A 3.82B parameter fine-tuned Phi-3-mini model specialized in structured information extraction with JSON templates and zero-shot capabilities.
Brief-details: Phixtral-4x2_8: A 7.81B parameter Mixture of Experts (MoE) model combining four Phi-2 variants, optimized for enhanced performance in text generation and coding tasks.
Brief Details: Specialized 7B parameter Mistral-based model focused on esoteric/occult knowledge. Fine-tuned on 10k GPT-4 synthetic instructions. Apache 2.0 licensed.
Brief Details: Japanese text-to-image diffusion model optimized for LORA compatibility with high-quality face generation and natural language prompts
Brief-details: A powerful 70B parameter LLM based on Llama2 that achieved breakthrough performance, being the first to surpass GPT-4 on AlpacaEval with 95.57% win rate vs Davinci-003.
Brief Details: Whisper-medium is a 769M parameter ASR model trained on 680k hours of data, supporting 99 languages with strong transcription and translation capabilities.
Brief Details: A powerful 3B parameter Voice-to-Voice model supporting 8 languages, with 700,000 hours of training data and semantic-token-free architecture.
Brief Details: A 70B parameter LLM fine-tuned from Llama-3, optimized for role-playing while maintaining strong general instruction-following capabilities.
BRIEF-DETAILS: ESPnet2 VITS-based TTS model trained on LJSpeech dataset, offering English text-to-speech capabilities with 2.7K+ downloads and CC-BY-4.0 license.
Brief-details: State-of-the-art open-source function calling model. 7B parameters, matches GPT-4 performance (83.67% accuracy). Multi-language support for Python, Java, JS & REST APIs.
Brief Details: Dual-purpose Stable Diffusion model offering both high-quality realistic and anime generation capabilities with optimized VAE settings.
BRIEF DETAILS: Qwen1.5-72B-Chat is a powerful 72.3B parameter language model featuring multilingual support, 32K context length, and improved chat capabilities with BF16 precision.
Brief Details: A 7B parameter Mamba-architecture LLM achieving competitive performance without attention mechanisms, trained on 5.5T tokens with strong math and reasoning capabilities.
Brief-details: Specialized Stable Diffusion model fine-tuned on Archer TV show screenshots. Features unique "archer style" artistic rendering with 216 likes and 1.1K+ downloads.
BRIEF-DETAILS: Microsoft's 7B parameter research model focused on reasoning capabilities, built on LLaMA-2 architecture. Excels in single-turn responses and comprehension tasks.
Brief-details: A fine-tuned 7B parameter LLaMA2 model optimized for Chinese language tasks, featuring enhanced Chinese dialogue capabilities through LoRA fine-tuning and comprehensive question-answering abilities.
Brief-details: SQLCoder-70B-Alpha is a powerful 69B parameter LLM fine-tuned from CodeLlama-70B, specializing in SQL generation with superior performance vs GPT-4.
Brief Details: SDXL-based ControlNet Tile model optimized for realistic image enhancement and detail boosting, featuring improved object recognition and color handling.
Brief Details: An open-source text-to-speech system built by inverting Whisper, supporting multiple languages and voice cloning with high-quality output at 1.5kbps.
