Claude Mythos

Claude Mythos, also designated Claude Mythos 5, is a language model from Anthropic, first disclosed in a company blog post on March 27, 2026. On the agentic_coding benchmark, which evaluates autonomous software engineering tasks, the model posted a score of 88.0, marking a significant leap in language model coding capabilities at the time of its emergence. Anthropic published this single benchmark result without releasing the parameter count, training data cutoff, or the score of its predecessor model on the same test. The March 27 blog post remains the sole official source confirming the model’s existence and its 88.0 agentic_coding figure, giving researchers and industry watchers a dated, verifiable data point for tracking progress in autonomous coding within a major foundation-model developer’s lineup. This entity matters because the 88.0 result provides a concrete, company-attested metric from early 2026 that can be used to calibrate the pace of improvement in AI-driven software development, directly influencing enterprise adoption trajectories and competitive dynamics, absent broader technical disclosures.