Brief Details: A 3B parameter vision-language model designed for web browsing automation, built on Qwen2.5-VL-3B-Instruct with 72.4% accuracy on WebVoyager benchmark.
Brief Details: NVIDIA's quantized version of DeepSeek R1, optimized for efficient inference with FP4 precision and 128K context length, running on TensorRT-LLM.
Brief Details: DeepSeek-V3 is a 671B-parameter MoE model (37B active) with 128K context, achieving SOTA performance in math/code tasks and supporting commercial use
Brief-details: Unified multimodal AI model leveraging decoupled visual encoding for both understanding and generation tasks, built on DeepSeek-LLM-7b-base with SigLIP-L vision capabilities.
Brief-details: A powerful 7B-parameter vision-language model capable of understanding images, videos, and structured data. Features enhanced visual comprehension, agent-like capabilities, and video analysis.
Brief-details: A 1.5B parameter distilled model from DeepSeek-R1, focused on reasoning capabilities, built on Qwen2.5-Math architecture with strong performance in mathematical and logical tasks.
Brief Details: Microsoft's phi-4: 14B parameter model with 16K context, trained on 9.8T tokens. Optimized for reasoning and safety, MIT licensed.
Brief-details: Meta's latest 70B parameter instruction-tuned LLaMA model, built for advanced natural language understanding and generation tasks with enhanced instruction-following capabilities.
Brief-details: Educational notebook collection demonstrating AI agent creation from scratch, part of Hugging Face Agents Course. Features practical examples and implementation guides.
BRIEF-DETAILS: 7B parameter Arabic language model from CohereForAI, focused on command-based tasks and language understanding, with 2025 version release
Brief-details: TinyR1-32B-Preview is a 32B parameter reasoning-focused model that achieves near-R1 performance in math, code, and science tasks through specialized domain training and model merging.
Brief Details: QwQ-32B-AWQ is a 4-bit quantized reasoning-focused language model with 32.5B parameters, featuring 131K context length and advanced architecture for enhanced problem-solving capabilities.
Brief-details: Wan2.1-T2V-1.3B is a lightweight (1.3B parameters) text-to-video generation model running on consumer GPUs, capable of creating 480P videos with only 8.19GB VRAM usage.
BRIEF-DETAILS: Stable Diffusion 3.5 Large - Latest advanced text-to-image model from Stability AI with enhanced capabilities and improved image generation quality
Brief Details: NotaGen - A symbolic music generation AI model using LLM training paradigms, featuring 3-stage training on 1.6M pieces and CLaMP-DPO optimization.
Brief Details: NeoBERT - A 250M parameter next-gen BERT model trained on RefinedWeb, featuring 4096 token context length and state-of-the-art MTEB benchmark performance.
BRIEF-DETAILS: Advanced UI screenshot parsing tool that converts interface elements to structured format. Features improved latency (0.6s/frame on A100) and 39.6 accuracy on ScreenSpot Pro.
BRIEF-DETAILS: Chroma: 8.9B parameter rectified flow transformer for text-to-image generation, built on FLUX.1 with architectural enhancements
Brief-details: QwQ-32B-GGUF is a powerful 32.5B parameter reasoning model from Qwen, featuring 131K context length and advanced architecture with RoPE, SwiGLU, and RMSNorm.
Brief-details: 8B parameter diffusion model trained from scratch, designed for instruction-following tasks. Comparable to LLaMA3 8B performance.
Brief-details: DiffRhythm-base is a pioneering diffusion-based song generation model capable of creating full-length songs in just 1m35s, using latent diffusion techniques.
