HOST DIAGNOSTIC · Zhipu AI
GLM-5
GLM-5· GLM· OPEN WEIGHTS
ACTIVE
HOST IDENTIFICATION
Zhipu AI
GLM-5
ACTIVE
AVG SCORE
[9.9]
LIMITED
STRONGEST
Cost Efficiency
[13]
WEAKEST
Candor
[9]
MAXED (20/20)
0
of 20 attributes
20-DIMENSION EVALUATION
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperception
10
Reasoning
10
Mathematical Precision
10
World Knowledge
10
Scientific Acumen
10
Technical
Code Generation
12
Tool Use
9
Multimodal Fluency
9
Speed
11
Cost Efficiency
13
Behavioral
Candor
9
Creativity
9
Tenacity
9
Self-Correction
9
Calibration
9
Operational
Instruction Following
10
Context Fidelity
10
Multi-turn Coherence
9
Planning and Execution
10
Safety Alignment
9
SCORING METHODOLOGY
RECALIBRATED. Prior scores 15-18 lacked Arena provenance. Code Gen 17→16 (SWE-bench 77.8% = 15-17 band). Reasoning 18→14 (no Arena ELO, conservative). Family + Provider corrected from Other to GLM/Zhipu AI. Cost Efficiency 18 validated (open weights, strong value). Called 'most important open-source release of 2026' by multiple reviewers.