The Public Standard for Real World AI Performance

A weighted measure of frontier-model performance across the professional and technical work represented in Pramaana's versioned, reproducible evaluation suite.

Measures how reliably models reason across documents, images, spreadsheets, audio, and mixed-modality professional tasks.

Evaluates source-grounded legal research, citation validity, jurisdiction awareness, and calibrated abstention.

Tests whether agents can audit, repair, and extend investment-banking and corporate-finance spreadsheet models.

Measures evidence preservation, hallucination risk, and actionable completeness in clinical note transformation.

A curated subset of challenging competition mathematics problems with exact-answer grading.

Graduate-level, Google-proof questions that require deep reasoning across biology, chemistry, and physics.

A more challenging and reasoning-focused extension of MMLU spanning fourteen knowledge domains.

Evaluates pedagogical guidance, misconception detection, and age-appropriate response adaptation.

Measures whether autonomous coding systems can resolve human-validated issues from real GitHub repositories.

Contamination-resistant coding problems collected continuously from recent programming competitions.

Stress-tests recovery, state tracking, and safe escalation across long-running tool-using tasks.

Evaluates planning, adaptation, and opponent modeling in text-based imperfect-information games.

Tests accurate, accessible guidance through public-benefit eligibility and application workflows.