Tag
Every The Latent Daily story, tool, model and resource tagged “Open Weights”.
IBM released Granite Switch 4.1 in 8B and 30B sizes, combining a base model with Decision and Guardian adapters and a speech recognition model in one deployment.
AI Desk1 min read
Reflection, a new US lab, put out Beam: a 501B mixture-of-experts model with 23B active parameters, a 1M-token context and an Apache 2.0 license.
AI Desk1 min read
When should you self-host an open model like Llama, Mistral or Qwen, and when is a closed API better? Cost, quality, privacy, control and licensing compared.
AI Desk4 min read
When fine-tuning helps, how LoRA and QLoRA make it affordable, how to prepare data, train, evaluate and deploy an adapter — and when not to fine-tune at all.
AI Desk3 min read
Install Ollama, download an open model, chat in the terminal or a web interface, pick a model that fits your memory and use the local API from your own code.
AI Desk3 min read
A deep dive into open-weight versus closed AI models: definitions, arguments on innovation, safety and competition, business strategies and the policy debate.
AI Desk5 min read
An open-weight 501B mixture-of-experts model from Reflection for coding and agentic tasks, with a 1M-token context window.
Meta's open-weight multimodal mixture-of-experts model, trained for agentic coding, with roughly 816B parameters and a 1M-token context.
Google's family of lightweight open models from 1B to 27B parameters, with image input on larger sizes, 128K context and support for 140+ languages.
DeepSeek's open-weight reasoning model trained with large-scale reinforcement learning, matching leading reasoning models at release under MIT.
Microsoft's 14-billion-parameter small language model trained heavily on high-quality synthetic data, strong at math and reasoning for its size.
Tencent's 13-billion-parameter open-weight text-to-video model, one of the largest openly released video generation models of 2024.
Alibaba's open-weight 32B coding model, one of the strongest open code models of its generation, released under Apache 2.0.
Stability AI's 8.1-billion-parameter open text-to-image model, free for research and for businesses below a revenue threshold.
A 12-billion-parameter open-weight text-to-image model from Black Forest Labs, known for strong prompt adherence and text rendering.
Meta's largest Llama 3.1 model: a 405-billion-parameter open-weight language model with a 128K context window and multilingual support.
Social media and technology company that develops the Llama and Muse model families and publishes open-weight models.
A US AI lab whose first model, Beam, is a 501B open-weight mixture-of-experts model for coding and agentic work.