CAIS and Scale AI release Humanity's Last Exam (3,000-question public release)

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Released Humanity's Last Exam (public release of 3,000 questions)

Why it matters

The next frontier after MMLU saturation; models at the time scored under 10%

Source — Primary sources first; secondary reports are labelled “Press”

Officialscale.comhttps://scale.com/blog/humanitys-last-exam-results

Added in Week 35, 2026 · See that week

On the map — ⑥ Reasoning Models line

  1. Previous stopDeepSeek-R1
  2. This recordCAIS and Scale AI release Humanity's Last Exam (3,000-question public release)
  3. Next stopGPT-4.5
See it on the map

Same week

OpenAI releases Operator and the CUA modelOfficialWhite House signs Executive Order 14179GovernmentEpoch AI posts a clarification on how FrontierMath was made and is usedOfficialDeepSeek-R1Paper

Nearby on the same axis

DeepSeek-R1PaperNvidia loses about $589 billion, the biggest single-stock drop in US historyPressFrançois Chollet leaves Google and founds Ndea, a lab dedicated to AGIOfficialOpenAI releases o3-miniOfficial