Rankings

MHIL
Leaderboard

The MHIL leaderboard will publish results as evaluations are completed. The table below is an illustrative projection showing how leading AI systems are expected to perform under the MHIL cognitive framework, based on design analysis and publicly available benchmark data.

The MHIL leaderboard is launching with initial evaluations. The projected scores below illustrate how systems are expected to score under the MHIL framework. No system is projected to exceed MHIL-65. For reference, MHIL-100 represents average adult cognitive performance.

ILLUSTRATIVE PROJECTION | EVALUATIONS IN DEVELOPMENT

12 SYSTEMS PROJECTED

#	System	MHIL Score	Strongest Dim.	Weakest Dim.	Tier
1	Anthropic ░░░░░░ ██ 3.█	62	Logical Reasoning	Metacognition	ADVANCED
2	xAI ░░░░ █.█	58	Critical Thinking	Metacognition	BASIC
3	Google Deep░░░░ ██ 2.█	56	Working Memory	Creative Prob-Solving	BASIC
4	OpenAI ░░░░ █-██████	54	Social Intelligence	Metacognition	BASIC
5	Meta ░░░░ ███-█ 4██	51	Logical Reasoning	Adaptive Behavior	BASIC
6	Mistral ░░░░░ ██████	49	Logical Reasoning	Causal Understanding	BASIC
7	Cohere ░░░░░░ ██ R+	46	Working Memory	Metacognition	BASIC
8	Amazon ░░░░░░ ████	44	Critical Thinking	Creative Prob-Solving	BASIC
9	Alibaba ░░░░ █.█-███	43	Logical Reasoning	Social Intelligence	BASIC
10	░░░░░░░░ ██ Labs v█	41	Working Memory	Metacognition	BASIC
11	░░░░░░ AI ████████	38	Social Intelligence	Causal Understanding	BASIC−
12	░░░░░░░ ██████ 1.█	34	Logical Reasoning	Metacognition	BASIC−

Illustrative projected scores based on design analysis. System identities partially redacted. Full results will be published as evaluations are completed. Human adult baseline: MHIL-100.

Projected Observations

Key Projections

All Systems Projected Below MHIL-65

Based on design analysis, no system is projected to reach adolescent-level cognition. The best-performing system is projected at 62 — equivalent to a cognitively advanced child, not an adult.

Metacognition Expected to Be Universally Weak

The framework is designed to expose this gap: current AI systems are expected to score below MHIL-35 in metacognitive awareness — fundamentally unable to assess what they do and do not know.

Fluency Masks Cognitive Deficits

Systems with high social intelligence scores create an illusion of competence. Sophisticated language output is not evidence of sophisticated thinking — a core design insight of the MHIL framework.

aggregate_projection.dat

AGGREGATE DIMENSIONAL PROJECTION — ILLUSTRATIVE // Projected average scores across 12 systems ─────────────────────────────────────────────── Logical Reasoning: 56 ███████░░░ Child-Adolescent Social Intelligence: 53 ██████░░░░ Child Working Memory: 49 ██████░░░░ Child Critical Thinking: 47 █████░░░░░ Child Causal Understanding: 42 █████░░░░░ Child Adaptive Behavior: 38 ████░░░░░░ Child Creative Problem-Solving: 34 ████░░░░░░ Child Metacognitive Awareness: 23 ██░░░░░░░░ Minimal ─────────────────────────────────────────────── MEAN PROJECTED: 42.8 (Child-level cognition) // Human adult baseline: 100 // Gap to close: 57.2 points // No system within 35 points of adult cognition

Interested in Early Collaboration?

Researchers and AI developers interested in early access or collaboration can get in touch.