ColabFold is a software tool that accelerates the prediction of protein structures and complexes by combining the fast homology search capabilities of MMseqs2 with AlphaFold2 or RoseTTAFold. By replacing AlphaFold2's homology search with MMseqs2, ColabFold achieves a 40-60-fold faster search and optimized model utilization, enabling the prediction of nearly 1,000 structures per day on a server with one graphics processing unit (GPU). Coupled with Google Colaboratory, ColabFold becomes a free and accessible platform for protein folding. It consists of three parts: an MMseqs2-based homology search server, a Python library for preparing input features and visualizing results, and Jupyter notebooks for basic, advanced, and batch use. ColabFold outperforms AlphaFold-Colab and matches AlphaFold2 on CASP14 targets and AlphaFold-multimer on the ClusPro dataset in prediction quality. It also offers advanced features such as sampling diverse structures, custom MSAs, and a lightweight 2D structure renderer. Benchmarking with CASP14 targets and a proteome benchmark of *Methanocaldococcus jannaschii* demonstrate its accuracy and speed. ColabFold is open-source software available at <https://github.com/sokrypton/ColabFold> and its environmental databases are available at <https://colabfold.mmsseqs.com>.ColabFold is a software tool that accelerates the prediction of protein structures and complexes by combining the fast homology search capabilities of MMseqs2 with AlphaFold2 or RoseTTAFold. By replacing AlphaFold2's homology search with MMseqs2, ColabFold achieves a 40-60-fold faster search and optimized model utilization, enabling the prediction of nearly 1,000 structures per day on a server with one graphics processing unit (GPU). Coupled with Google Colaboratory, ColabFold becomes a free and accessible platform for protein folding. It consists of three parts: an MMseqs2-based homology search server, a Python library for preparing input features and visualizing results, and Jupyter notebooks for basic, advanced, and batch use. ColabFold outperforms AlphaFold-Colab and matches AlphaFold2 on CASP14 targets and AlphaFold-multimer on the ClusPro dataset in prediction quality. It also offers advanced features such as sampling diverse structures, custom MSAs, and a lightweight 2D structure renderer. Benchmarking with CASP14 targets and a proteome benchmark of *Methanocaldococcus jannaschii* demonstrate its accuracy and speed. ColabFold is open-source software available at <https://github.com/sokrypton/ColabFold> and its environmental databases are available at <https://colabfold.mmsseqs.com>.