Open-Weight Models

Mistral says its new open-weight Le Chonk model is the strongest offering outside China, raising fresh questions about performance and deployment.

Aleph Alpha released Kolibri, an open-weight English-German MoE with 3.46B active parameters and a reported 1M context window for local AI deployment.

AWS shows developers how to pair OpenCode with open-weight models in Amazon Bedrock, bringing private, flexible, pay-per-use AI coding agents to AWS.

AWS has added Moonshot AI’s Kimi K3 to Amazon Bedrock, giving builders an open-weight model with vision, long context, and prompt caching.

Base Labs, Hugging Face, and Goodfire are building open-weight AI safety infrastructure, aiming to make safeguards more transparent as model risks grow.

A Mozilla report says China’s open-weight AI models are 4.4 months behind US frontier systems, while lower costs could speed adoption for builders and enterprises.

Arcee AI has reached a valuation above $1 billion in a Series B round, underscoring investor interest in open-weight models for enterprise AI.

An AA founder says South Korea’s open-weight AI and manufacturing advantages could support global competition, but warns the country cannot lose ground in AGI.

The Institute of Foundation Models has released K2 Horizon models with weights, code and training data, expanding access to inspectable AI development.

Y Combinator CEO Garry Tan backs an American regime allowing open-weight labs to distill frontier models, challenging restrictions and raising policy questions.

Abacus.AI has launched three open-weight Smaug models for enterprise AI agents, adding a new option for teams building agentic workflows.

MBZUAI’s Institute of Foundation Models says K2 Horizon opens weights, code, data and methods, challenging closed AI on reproducibility.

Reuters has flagged an open-weight AI licensing development, but the available report does not identify the model, owner, terms, or business impact.

Tencent’s reported Hy3 model is described as a 770B open AI system for coding, but limited source evidence leaves its release and capabilities unverified.

Reported bids for Hugging Face, Poolside, and OpenRouter show why major technology companies are pursuing open-weight AI infrastructure and talent.

Meta has released Muse Glimmer, a 30B open-weight multimodal model for local agents, coding, and private image and video workflows on varied hardware.

Perceptron, founded by former Meta researchers, launches open-weight Isaac 0.5 to give industrial robots flexible visual reasoning and control.

The Hims CEO says companies with large datasets could reduce AI spending by up to 80% with open-weight models, raising key deployment questions.

Z.ai’s GLM-5.3 ties for the top open-model score while cutting task costs versus Kimi K3, but open weights face a security delay.

Reports link Alibaba to a laptop-ready open-weight model as Meta unveils Muse Glimmer, sharpening competition over local AI and developer control for builders.