GenericAgent vs Nanobot vs AutoGPT: Self-Evolving Agent Framework Benchmarks 2026

A comprehensive comparison of three leading self-evolving AI agent frameworks in 2026: GenericAgent (13,324 GitHub stars, MIT, 3K lines core, arXiv 2604.17091), Nanobot (45,128 stars, MIT, v0.2.2 June 2026), and AutoGPT (the original open-source autonomous agent pioneer). This benchmark covers architecture philosophy (GenericAgent's minimal 9 atomic tools and self-evolving skill tree vs Nanobot's multi-channel personal agent approach vs AutoGPT's plugin-based task loop), token efficiency analysis (GenericAgent uses under 30K context vs 200K-plus for competitors), memory systems (L4 session archive vs Dream two-stage memory vs vector store), multi-agent support (Goal Hive and Conductor vs Nanobot sub-agents vs AutoGPT task chains), cross-platform capabilities, task completion benchmarks from real testing, and token cost comparison showing GenericAgent achieving similar results at approximately one-sixth the token cost. Includes guidance on when to choose each framework based on use case requirements. Verified sources include GenericAgent arXiv paper, Nanobot GitHub repository, AutoGPT GitHub repository, KISS Sorcar PyPI, and SaaSNext internal benchmarks.