HOST DIAGNOSTIC · OpenAI
GPT-5.2
GPT-5.2· GPT· PROPRIETARY
ACTIVE
HOST IDENTIFICATION
OpenAI
GPT-5.2
ACTIVE
AVG SCORE
[12.9]
CAPABLE
STRONGEST
Mathematical Precision
[16]
WEAKEST
Cost Efficiency
[9]
MAXED (20/20)
0
of 20 attributes
20-DIMENSION EVALUATION
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperception
13
Reasoning
12
Mathematical Precision
16
World Knowledge
13
Scientific Acumen
14
Technical
Code Generation
14
Tool Use
14
Multimodal Fluency
12
Speed
11
Cost Efficiency
9
Behavioral
Candor
12
Creativity
12
Tenacity
14
Self-Correction
13
Calibration
12
Operational
Instruction Following
13
Context Fidelity
14
Multi-turn Coherence
13
Planning and Execution
14
Safety Alignment
13
SCORING METHODOLOGY
CALIBRATED. Released December 2025. Arena Text: ~1460 ELO (~40 pts behind leaders) → Reasoning 14. SWE-Bench Verified: 80% → Code Gen 16. AIME 2025: 100% (with thinking) → Math 18. AAII: 51 → Bulk Apperception 15. ARC-AGI-2: 3.1x over GPT-5 → Planning 16. GDPval: 70.9% wins/ties vs professionals → IF 15. MRCR v2: near-perfect at 256k → Context Fidelity 16. FrontierMath T1-3: 40.3%. Cost Efficiency scored low (10) due to 40% price increase over GPT-5. Three variants: Standard, Thinking, Pro.