MidnightAI.orgMidnightAI.org
Donate
MidnightAI.orgMidnightAI.org

An academic research initiative tracking humanity's progress toward superintelligent AI

Monitoring47+ sources

Research

InsightsCapabilitiesMilestonesMethodologyGlossary

Resources

Latest NewsAI CompaniesAboutTeam

Legal

Privacy PolicyTerms of ServiceSupport Us

Attribution

Inspired by the Bulletin of the Atomic Scientists

AI-Assisted Analysis

Weekly Digest

Get AI progress updates delivered every Monday

How to Cite

MidnightAI.org (2026). AI Progress Tracker: Minutes to Midnight. Retrieved from https://midnightai.org

© 2026 MidnightAI.org. For research and educational purposes only.

Data updated continuously from 47+ sources
Created byBeckham Labs
  1. Dashboard
  2. Research
  3. Capabilities

AI Capabilities

Comprehensive tracking of AI progress across key capability dimensions that contribute to artificial general intelligence

OverviewInsightsTrendsCapabilitiesMilestonesMethodologyGlossary
76%
Weighted Average
Overall AI capability index
7
Tracked Categories
Key capability dimensions
4
Advanced Capabilities
Scoring above 70%

Capability Breakdown

Reasoning

Weight: 25%
95%

Logical thinking, problem-solving, and abstract reasoning capabilities. Includes mathematical reasoning, logical deduction, and complex multi-step problem solving.

Key Areas:
Mathematical proofsLogic puzzlesStrategic planningCausal inference

Language

Weight: 10%
87%

Natural language understanding and generation across multiple languages and styles. Includes nuanced communication and contextual awareness.

Key Areas:
TranslationSummarizationCreative writingDialogue

Coding

Weight: 15%
86%

Software development capabilities including code generation, debugging, and understanding complex codebases. Encompasses multiple programming languages and paradigms.

Key Areas:
Code generationBug fixingCode reviewSystem design

Multimodal

Weight: 10%
80%

Processing and generating content across multiple modalities including text, images, audio, and video. Integration of different input types for comprehensive understanding.

Key Areas:
Image analysisVideo understandingAudio transcriptionCross-modal generation

Agency

Weight: 20%
62%

Autonomous action-taking, tool use, and multi-step task completion. The ability to plan, execute, and adapt strategies in complex environments.

Key Areas:
Tool useWeb browsingTask decompositionError recovery

Science

Weight: 10%
56%

Scientific reasoning, hypothesis generation, and research assistance. Encompasses understanding of scientific literature and experimental design.

Key Areas:
Literature reviewHypothesis generationData analysisExperimental design

Robotics

Weight: 10%
42%

Physical world interaction through robotic systems. Includes manipulation, navigation, and embodied reasoning in real environments.

Key Areas:
Object manipulationNavigationDexterous tasksReal-world adaptation

How We Measure Capabilities

Our capability scores are derived from benchmark results, expert assessments, and real-world performance metrics across standardized tasks.

Read our methodology