Model training

Models 4

Gen 7

Apertus v1.1 4B

By Swiss AI Initiative

Apertus-v1.1-4B is a 4B parameter open-source base language model created via pre-training distillation from the Apertus-8B teacher model. It uses a dense transformer with grouped-query attention and xIELU activations, natively supports 1,811 languages, and is trained on 1.7T high-quality tokens. Designed for constrained hardware and further fine-tuning.

📚Large Language Models 🔍Research 🌐Multilingual communication 🧠Model training

NewText

Released 6d ago
Gen 7

nanowhale

By Hugging Face

nanowhale is Hugging Face’s miniature DeepSeek-V4-style language model project, built to reproduce the core DeepSeek-V4 architecture at about 110M parameters. It is designed as an educational and research-scale model, with both a pretrained base version and an SFT chat version, rather than as a production frontier assistant.

📚Large Language Models 🧠Model training 🦾Ai research

NewText

Released 2mo ago
Gen 3

davinci llm 3B

By Sii GAIR-NLP

daVinci-LLM-3B is a 3B base language model built to make pretraining transparent and reproducible. Its release includes not only the weights, but also training trajectories, intermediate checkpoints, data-processing decisions, and more than 200 ablation studies.

📚Non-interactive language modeling 🤖Ai research assistance 🔍Data quality control 🧠Model training

Text

Released 3mo ago
Gen 7

WebWorld 8B

By Alibaba

WebWorld-8B is Qwen’s open web-agent model for autonomous browser interaction and online task completion. It is built to navigate websites, interact with web interfaces, reason over browser state, and execute multi-step online workflows using an 8B parameter language-model backbone.

🌐Web browsing 🔀Synthetic data 🧠Model training

Text

Released 4mo ago