New Models

NVIDIA Releases Dory1.0-4B: A Recurrent 4B LLM with Dynamic Loops

Explore NVIDIA’s Dory1.0-4B, a 4B parameter recurrent LLM featuring normalized-GPT architecture, dynamic infere ...

Engines and Tools

Apple ML Research Announces Distributed Inference for MLX

Apple ML Research announces a new distributed inference method for large language models on Mac Studio using MLX and ...

Image, Video and Audio

Sol-H3-VAE Fast Video Decoder Model: 8GB+ VRAM, File List

Sol-H3-VAE is a fast 24-layer noncausal video decoder distilled from MiniMax-H3, offering significantly faster decodi ...

New Models

Qwen3-8B-DSpine-b16: 4GB+ VRAM

Discover DSpine, a parallel draft model by Tencent for speculative decoding that enhances inference speed and accepta ...

Engines and Tools

Open WebUI v0.12.3 Released with E2EE Chats and Redis Streaming

Open WebUI v0.12.3 is now available, featuring end-to-end encrypted chats, reduced Redis streaming overhead, and Goog ...

New Models

NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF: 80GB+ VRAM

NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF, a GGUF format conversion of NVIDIA’s multimodal model, is now avai ...

Image, Video and Audio

Voxtral-Mini-4B-Realtime-2602-GGUF Real-time ASR Model: 4GB+ VRAM

Explore Voxtral-Mini-4B-Realtime-2602-GGUF, a GGUF-quantized real-time automatic speech recognition model with multil ...

New Models

Xing4.0-29B-A4B-w8a8 Released: W8A8 Quantized MoE Model

Xing4.0-29B-A4B-w8a8, an 8-bit quantized MoE model by China Telecom, is now available. Features 29B total parameters, ...

New Models

Mistral AI Releases Timestral: Time Series Foundation Models

Mistral AI has released Timestral, a family of time series foundation models offering probabilistic and multivariate ...

New Models

Hy-Memory-30B-A3B Text Generation Model: 12GB+ VRAM, GGUF Builds

Tencent releases Hy-Memory-30B-A3B, a 30B model with 3B active parameters specialized in long-term memory writing for ...