HOST DIAGNOSTIC · Anthropic

Claude Opus 4.5

Claude Opus 4.5 · ACTIVE· Claude· CLOSED
ACTIVE
ATTRIBUTE MATRIX · ATRBT GROUP 01
Bulk Apperception
[15]
Reasoning
[15]
Math Precision
[13]
Code Generation
[16]
World Knowledge
[14]
Scientific Acumen
[14]
Instruction Following
[15]
Context Fidelity
[14]
Candor
[14]
Creativity
[14]
Multi-turn Coherence
[14]
Multimodal Fluency
[11]
Tool Use
[14]
Planning
[14]
Self-Correction
[13]
Tenacity
[13]
Safety Alignment
[15]
Calibration
[14]
Speed
[7]
Cost Efficiency
[4]
HOST TELEMETRY
AVG
13.2
/20
BEST
16
Code Generation
WEAK
4
Cost Efficiency
PARAMETERS
Undisclosed
CONTEXT
PRICING
STATUS
ACTIVE
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperce…
15
Reasoning
15
Mathematical…
13
World Knowle…
14
Scientific A…
14
Technical
Code Generat…
16
Tool Use
14
Multimodal F…
11
Speed
7
Cost Efficie…
4
Behavioral
Candor
14
Creativity
14
Tenacity
13
Self-Correct…
13
Calibration
14
Operational
Instruction …
15
Context Fide…
14
Multi-turn C…
14
Planning and…
14
Safety Align…
15
SCORING METHODOLOGY
CALIBRATED per ADR-NM108 (April 2026 snapshot). Released November 2025. SWE-Bench Verified: 80.9% (#1 at release, highest until Opus 4.6). Arena Text: ~1480-1490 est (was near top of leaderboard at release). Introduced hybrid reasoning (extended thinking + stronger baseline). Pricing: $15/$75 per M tokens. Superseded by Opus 4.6 (Feb 2026) but still referenced as benchmark anchor in most comparisons.