Falcon 180B released by TII on HuggingFace — open model with hosting restrictions

AI Impact Summary

TII releases Falcon 180B on HuggingFace, the largest openly available LLM at 180B parameters, trained on 3.5 trillion tokens with RefinedWeb across up to 4096 GPUs using SageMaker. The model and its chat variant are accessible on the Hugging Face Hub and Falcon Chat Demo Space, with tooling support via Transformers 4.33 and optimization options such as bitsandbytes for 8-bit/4-bit quantization and PEFT/GPTQ. Commercial use is allowed but under restrictive conditions that exclude hosting, requiring license review for production deployments.