Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossaryModelsIntermediate

Open weight

Also written: open weight model, Open Source Model, Open Source AI, Open-weight AI, Open-weight AI models, Open-weight LLMs, publicly released models, Open-weight cyber models

An AI model with publicly released weights, allowing anyone to use, modify, and build upon it.

In plain terms

An open weight model is an artificial intelligence system where the internal parameters, often called weights, are made public. This means researchers and developers can download, inspect, and even change the model's core workings.

Think of it like getting the engine of a car, not just the car itself. You can see how the engine is built, tune it for better performance, or even put it into a different car. This openness allows for faster innovation and wider use of AI technology, as people don't have to build these complex models from scratch.

The expert version

An open weight model refers to a machine learning model where the trained parameters (weights) are publicly released. This contrasts with closed models where these weights are proprietary. The release typically includes the weights and inference code, but may not include the full training dataset or methodology.

This approach fosters transparency, reproducibility, and community-driven development. It allows researchers to scrutinize model behavior, adapt models for specific tasks, and build upon existing work. However, it also raises concerns about potential misuse and the difficulty of controlling downstream applications.

Why it matters now

Open weight models are a major focus as companies like Meta and Google release powerful new versions. This trend is driven by the desire to enable developers to build and customize AI agents, as well as manage the associated risks in enterprise settings. The availability of these models also fuels innovation in areas like on-device AI and advanced tool use.

Where it shows up 49 items, newest first
How each company uses it
Anthropic Claude
Anthropic has released an open weight model named Holo4, which demonstrates capabilities in interacting with graphical user interfaces, coding, and tool usage.
Google Gemini
Google is making its AI models, including the Gemma series, available as open weight, allowing developers to build applications that can run on devices.
Meta Meta AI
Meta is actively releasing open weight models, such as its new large coding model, and partnering with others like Apple to bring these models to platforms like Macs.
Microsoft Copilot
Microsoft is addressing the enterprise side of open weight models by providing frameworks to manage the risks associated with their deployment.
Related terms

open source AImodel weightsfine tuningproprietary modelLLMparameter efficient fine tuning

Often confused with
Open Source AI
Open source refers to software where the source code is freely a
Publicly Available Models
This is a broader term that could include models acce
Research Models
Some models are released for research purposes but may have res
Origin

The concept of releasing model weights emerged from the machine learning research community, gaining significant traction with the open sourcing of models like LLaMA by Meta. Source ↗

Try this about 5 minutes

Find an open weight model, like Meta's Llama 2 or Google's Gemma, on a platform like Hugging Face. Download a smaller version and try running it locally using a simple interface or script to see how it responds to prompts. You can then experiment with changing a few parameters in the inference code to observe how it affects the output.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

What are model weights?
Model weights are the numerical parameters that an AI model learns during training. They essentially represent the 'knowledge' the model has acquired and determine how it processes input to produce output.
Can anyone use an open weight model?
Generally, yes, but the specific license under which the model is released dictates the terms of use, modification, and distribution. Some licenses may have restrictions on commercial use.
What's the difference between open weight and open source?
Open source typically refers to the availability of source code, while open weight specifically means the trained parameters of the AI model are publicly released. An open weight model is often part of a larger open source project.

First seen on Subvolts Aug 5, 2025. Updated Oct 18, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.