Benchmarks

Benchmarks

Our benchmarks measure abilities that matter for human understanding. Every evaluated reply and grader mark is published with the scores.

ReleasedOctober 2026

SENSUS Bench v1.0

How well a model follows a person's reasoning: whether its reply names the likely error, stays appropriately tentative, explains the idea and invites the next step. 76 evaluation cases in mathematical reasoning, eight models, two model graders.

Top models

  1. Qwen3.8-27B (LORM)0.838
  2. Qwen3-14B (LORM)0.720
  3. Qwen3-8B (LORM)0.626
Close window

Contact Us

20 to 6,000 characters.

Close window Close window