Meituan/Text / LLM
Meituan Ships a Lighter, Sparser LongCat-Flash
The food-delivery giant's newest open model trims its mixture-of-experts design for more efficient inference under an MIT license.
The feed
Latest releases
Every new open-source model release and major update — aggregated from across the ecosystem, deduplicated, and refreshed every 12 hours.
341 releases
DeepSeek/Text / LLMMajor release
DeepSeek Refreshes V4-Flash With New 0731 Checkpoint
The MIT-licensed mixture-of-experts model returns in an updated build shipping with FP8 weights for cheaper inference.
DeepSeek/Text / LLM
DeepSeek Ships V4-Flash, a 304B MoE Tuned for Agents
The latest checkpoint in DeepSeek's V4 line leans into agentic workflows while keeping the permissive MIT license.
LGAI EXAONE/Text / LLM
LG AI Research debuts K-EXAONE 2.0, a 750B MoE model
The new mixture-of-experts model activates 37B parameters per token and targets English, Korean, and Spanish reasoning tasks.
MiniMax/Text → Video
MiniMax Releases H3 Video Model on Hugging Face
The company's new diffusion model handles text-to-video and image-to-video, with support for joint audio-video generation.
Skt/Text / LLM
SK Telecom Releases A.X-K2 Multilingual LLM
The Korean telecom carrier's latest open language model targets English, Korean, Chinese, Japanese, and Spanish under a permissive license.
Audio8/Text → Speech
Audio8 debuts a 0.6B multilingual zero-shot TTS preview
The compact text-to-speech model promises voice cloning across languages from a footprint small enough to run without heavy hardware.
Thinkingmachines/Vision-Language
Thinking Machines Debuts Inkling Small, a Compact Multimodal MoE
The Apache-2.0 model brings mixture-of-experts efficiency to image, audio, and text tasks in a smaller footprint.
LiquidAI/Embeddings
Liquid AI's LFM2.5 encoder targets fast CPU inference
A 230M-parameter bidirectional encoder built for long-context English and German embeddings without a GPU.
LiquidAI/Embeddings
Liquid AI ships a 350M encoder built for CPUs
The compact LFM2.5 encoder targets fast, long-context text embeddings without a GPU.
KRAFTON/Any-to-Any
KRAFTON releases A.X-K2 Raon speech MoE model
The game maker's new open model blends text-to-speech and speech recognition in a single 21B mixture-of-experts system with just 3B active parameters.
Microsoft/Vision-Language
Microsoft's Mage-VL Streams Video Natively
A codec-native multimodal foundation model aims to understand live video and vision-language input in real time.
Swiss Ai/Text / LLM
Apertus v1.5 70B arrives with an Apache-2.0 license
Switzerland's open-model effort ships a 70-billion-parameter, multilingual and multimodal system that anyone can use, modify, and deploy.
Microsoft/Speech → Text
Microsoft's VibeVoice ASR Goes BitNet for CPU Speech
A BitNet-quantized speech recognition model trades GPU dependence for efficient CPU inference in English and Chinese.
Amd/Reasoning
AMD's Instella-MoE Brings Reasoning to ROCm Hardware
A new open mixture-of-experts model with 16B total parameters and just 3B active is tuned to run on AMD's own accelerator stack.
Kwaipilot/Code
Kwaipilot Releases KAT-Coder V2.5 Dev, an Agentic MoE Coder
Kuaishou's coding team ships an open mixture-of-experts model built on the Qwen3.5 MoE architecture and tuned for agentic development work.
Upstage/Text / LLM
Upstage's Solar Open2 arrives as a 250B MoE model
The Korean AI firm's latest open release scales to 250 billion parameters with a mixture-of-experts design tuned for English and Korean.
Microsoft/Text → Image
Microsoft's Mage-Flow packs image editing into 4B
A compact model handles both text-to-image generation and instruction-based edits at native resolution, under a permissive MIT license.
Motif Technologies/Text / LLM
Motif Technologies debuts Motif 3 Beta, an MoE model
The Korean AI lab's preview release is a mixture-of-experts language model built for long-context, multilingual work.
Microsoft/Vision-Language
Microsoft's Fara1.5-27B targets computer-use agents
A 27B-parameter vision-language model built to drive browsers and desktop apps like a human operator.
