Explore the architecture behind our models:Read our white paper →

Introducing MedLect Mini & MedLect Pro

MedLect Pro reasons through clinical cases linking every claim back to PubMed for full transparency.

Model Overview

  • MedLect Mini: High-speed flashcard generator with live source links.
  • MedLect Pro: Advanced clinical reasoning fine-tuned on 90,000+ expert Q&A pairs.
  • Real-time PubMed retrieval ensures every answer is evidence-based.

Why It Matters

MedLect Pro scores 91.2% on 5,000 unseen USMLE-style questions outperforming o3, Claude 3.7 Sonnet, and Gemini 2.5 Pro while maintaining 95% citation precision.

  • Handles edge-case clinical scenarios with minimal hallucinations.
  • Every answer links back to peer-reviewed sources.

Evidence-Based Learning

  • Built on Qwen-3 235B with LoRA adapters and a stepwise USMLE curriculum.
  • Curriculum-scheduled training across increasing difficulty tiers.
  • Hybrid BM25 + FAISS pipeline pulls from PubMed & Google Scholar.
  • 95% citation precision, verified against peer-reviewed literature.

Designed by Students, For Students

  • Refined through real study sessions and feedback loops.
  • Aligned with how medical students actually learn and retain information.

Proven Results

Independent evaluation shows MedLect Pro outperforms other LLMs in USMLE-style accuracy, with experts preferring our explanations 68% of the time.

LLM Benchmark Accuracy: MedQA, MedMCQA, PubMedQA (%)

MedLect Pro leads the pack with 91.2% accuracy on MedQA.

Scores represent model accuracy on three leading USMLE-style question benchmarks.