HOST DIAGNOSTIC · Google
Gemini 2.5 Deep Think
Gemini 2.5 Deep Think· Gemini· PROPRIETARY
ACTIVE
HOST IDENTIFICATION
Google
Gemini 2.5 Deep Think
ACTIVE
AVG SCORE
[12.5]
CAPABLE
STRONGEST
Reasoning
[15]
WEAKEST
Speed
[5]
MAXED (20/20)
0
of 20 attributes
20-DIMENSION EVALUATION
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperception
14
Reasoning
15
Mathematical Precision
15
World Knowledge
13
Scientific Acumen
15
Technical
Code Generation
13
Tool Use
11
Multimodal Fluency
12
Speed
5
Cost Efficiency
8
Behavioral
Candor
13
Creativity
12
Tenacity
13
Self-Correction
12
Calibration
13
Operational
Instruction Following
13
Context Fidelity
14
Multi-turn Coherence
13
Planning and Execution
14
Safety Alignment
12
SCORING METHODOLOGY
CALIBRATED. Deep reasoning variant of Gemini 2.5 family. ARC-AGI-2: 45.1% (Gemini 3 Deepthink variant). Math + Science scored high (17) as extended reasoning model. Speed scored low (implicit slow reasoning). Cost 9 (extended thinking burns tokens). On LM Council benchmark suite.