Thinking Machines Lab releases open-weight Inkling 2 Mini model that runs on a laptop
Thinking Machines Lab has released Inkling 2 Mini, an open-weight model built on a recurrent Transformer architecture…
The Open Source topic collects downloadable, self-hostable open-weight models and frameworks: trending GitHub projects, license changes, and community progress. Readers building their own inference services or studying model details can find entry points and docs here.
Open-weight models that can be downloaded and deployed independently, open-source frameworks, and AI projects trending on GitHub.
Thinking Machines Lab has released Inkling 2 Mini, an open-weight model built on a recurrent Transformer architecture…
Tencent Hunyuan has released the open weights of Hy4, positioned as a native multimodal flagship MoE model with 770B…
Browser Use releases browser-use 1.0, replacing the 24 browser actions of 0.x with a single persistent python REPL…
Ai2 on November 19 launched Theorizer on its research agent platform Asta. The agent retrieves literature and proposes…
Tencent Hunyuan says Hy4 flash is now open-weight, including two versions: Hy4 flash-FP8 and Hy4 flash-Base. The model…
Nathan Lambert writes that Nvidia CEO Jensen Huang's resignation will hit the US open-source model ecosystem, arguing…
Hugging Face CEO Clément Delangue and @ggerganov announced the release of llama.cpp v1.0.0, the first fully stable…
Meituan's LongCat team released the open-source audio-driven human video generation model LongCat-Video-Avatar 1.6,…
Edge0 announced the open-sourcing of the Edge0-145B and Edge0-120B models, along with a unified inference framework…
Runway says its world action model Praxis-1 is now available to all as open weights under Apache 2.0. The model is a…
On 10 November 2026, n8n released Nod as a research preview: an Apache 2.0 personal AI agent, separate from n8n, that…
MiniMax released H3-Image, an open-weight image generation model based on H3, supporting text-to-image,…
n8n Labs released the research preview of Nod, an open-source personal agent app independent of n8n that runs locally…
At the CoRL conference, NVIDIA released the open world action model GR00T 2, trained on Cosmos and using 100,000 hours…
Meituan's LongCat has open-sourced LongCat-2.5 under the MIT license, supporting commercial deployment and further…
Zyphra has released ZONOS3-Mini, an Apache 2.0 open-source zero-shot voice cloning TTS model built on a sparse MoE…
Reflection published a blog post detailing the safety training and evaluation methods for its open-weight model Beam…
Moonshot AI released the weights of Kimi K3.1 Pro on Hugging Face, fulfilling its commitment made at the October 24…
The OpenClaw Foundation has released Multiplayer, allowing multiple people to operate the same agent simultaneously…
Preferred Networks released the translation-specialized LLM PLaMo 3 Translate, used in its own translation service, as…
Baidu's ERNIE team has released ERNIE-5.5-Flash-Base, opening the pretraining endpoint checkpoint, which has not…
Meituan's LongCat team has released LongCat-2.5, with 1.6 trillion total parameters and about 48 billion activated per…
Tencent Hunyuan released the official version of Hy4, expanding pretraining and post-training scale compared with Hy4…
The author open-sourced Agent-Autoloop, an agent automation system he refined over a year, at…
Mistral team member Pierre Stock announced that ML4 (Mistral Large 4) is now open-weight, saying it has frontier…
According to David Abitbol, head of AI operations at Revolut, the company migrated about 2,000 engineers to…
Tencent Hunyuan releases Hyra 2.0, shifting the focus of automated scientific research from directly solving problems…
Pika Labs post-trained on MiniMax H3 to launch and open-source Pika H3 Flex, cutting sampling steps from 50 to 6,…
Tencent Hunyuan has released the technical report for its open-source interactive video world model Hy-Nomas…
Unsloth released a Dynamic 1-bit GGUF of Mistral Le Chonk (Mistral Large 4, a 1T-parameter MoE with 52B active), which…
Hugging Face published a launch blog for Mistral Large 4 (Le Chonk), a multimodal multilingual MoE model with about 1T…
The vLLM team announced Day 0 support for Mistral's new flagship model Mistral Large 4, a 1T-parameter native…
OpenMOSS released MOSS-Image, an open-weight image generation model pretrained from scratch using a 5B-parameter…
ModelBest released and open-sourced the MiniCPM-OCR series. MiniCPM-OCR 1.0 (1.3B) achieved SOTA scores of 93.68 on…
LangChain details how its in-house design agent Content Studio generates code-based assets such as videos, slide decks…
Liquid AI released the next-generation LFM3 model family, built on the shared LFM3-1.7B-Base backbone. It initially…
Unsloth released quantized GGUF versions of MiniMax-M3.1-Flash; the 2-bit version retains 91.5% top-1 accuracy while…
On October 26, Tencent Hunyuan released and open-sourced the simultaneous interpretation large model Hy-SimulMT,…
Tencent Hunyuan released Hyra 2.0, shifting automated research from directly solving problems to having agents develop…
Moonshot AI released its flagship model Kimi K3.1 Pro with 2.8T parameters and support for a 1M token context; full…
Sponsored
Orange Relay (pro.nat6.net) is a premium AI API relay — an excellent transfer station with competitive pricing, reliable quality and broad model coverage including Claude, GPT and Gemini.
Visit Orange Relay