Ampersand Tutor: humans& AI tutor tops the TutorMoments benchmark
Why it matters
humans& released Ampersand Tutor (&Tutor), an AI model trained to teach, available via API. It practised on Persimmon, the humans& learner model, and was rewarded by how much the student learned. On TutorMoments (plain prompt) it scored 0.754 vs 0.667 for Claude Sonnet 5.5 and 0.634 for GPT-6 Astra; with an evaluation-aware prompt GPT-6 Astra edges it, 0.926 to 0.919. Scores use a simulated student and an LLM judge; real-world learning tests are still to come.
Read the source: humans& (tutor.humansand.ai, Oct 30, 2026) (opens in a new tab)
All hits from 31 Oct 2026More ModelsGet the daily hits by email