Profluent-E1 family of single-sequence and retrieval-augmented protein language models. Designed as drop-in replacements for ESM-family models, available in 150M, 300M, and 600M parameter variants. Supports masked-token prediction, sequence/residue embeddings, and zero-shot variant fitness scoring (wildtype-marginal or masked-marginal) in either single-sequence or retrieval-augmented (MSA-context) mode.