JOHNCLICK.AI
HOST DIAGNOSTIC · OpenAI

GPT-5.3 Codex

GPT-5.3 Codex · ACTIVE· GPT· CLOSED
ACTIVE
ATTRIBUTE MATRIX · ATRBT GROUP 01
Bulk Apperception
[14]
Reasoning
[14]
Math Precision
[14]
Code Generation
[17]
World Knowledge
[13]
Scientific Acumen
[13]
Instruction Following
[15]
Context Fidelity
[14]
Candor
[13]
Creativity
[11]
Multi-turn Coherence
[13]
Multimodal Fluency
[10]
Tool Use
[17]
Planning
[16]
Self-Correction
[14]
Tenacity
[15]
Safety Alignment
[13]
Calibration
[13]
Speed
[11]
Cost Efficiency
[8]
HOST TELEMETRY
AVG
13.4
/20
BEST
17
Code Generation
WEAK
8
Cost Efficiency
PARAMETERS
Undisclosed
CONTEXT
—
PRICING
—
STATUS
ACTIVE
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperce…
14
Reasoning
14
Mathematical…
14
World Knowle…
13
Scientific A…
13
Technical
Code Generat…
17
Tool Use
17
Multimodal F…
10
Speed
11
Cost Efficie…
8
Behavioral
Candor
13
Creativity
11
Tenacity
15
Self-Correct…
14
Calibration
13
Operational
Instruction …
15
Context Fide…
14
Multi-turn C…
13
Planning and…
16
Safety Align…
13
SCORING METHODOLOGY
CALIBRATED per ADR-NM108 (April 2026 snapshot). Released February 5, 2026. OpenAI's most capable agentic coding model. Moves beyond writing code to using code as a tool to operate computers end-to-end. Arena Code: ~1490-1510 est. SWE-Bench Pro: competitive with GPT-5.4. Code Gen scored high (17) as a dedicated coding agent. Planning+Execution scored high (16) for agentic capabilities. Creativity scored low (11) as a coding specialist, not a generalist.
← ALL HOSTS
COMPARE IN MATRIX →KNOWLEDGE GRAPH →