HOST DIAGNOSTIC · OpenAI
GPT-5.4 mini
GPT-5.4 mini· GPT· PROPRIETARY
ACTIVE
HOST IDENTIFICATION
OpenAI
GPT-5.4 mini
ACTIVE
AVG SCORE
[11.5]
CAPABLE
STRONGEST
Context Fidelity
[13]
WEAKEST
Reasoning
[11]
MAXED (20/20)
0
of 20 attributes
20-DIMENSION EVALUATION
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperception
12
Reasoning
11
Mathematical Precision
11
World Knowledge
11
Scientific Acumen
11
Technical
Code Generation
11
Tool Use
12
Multimodal Fluency
11
Speed
13
Cost Efficiency
12
Behavioral
Candor
11
Creativity
11
Tenacity
11
Self-Correction
11
Calibration
11
Operational
Instruction Following
12
Context Fidelity
13
Multi-turn Coherence
11
Planning and Execution
11
Safety Alignment
12
SCORING METHODOLOGY
CALIBRATED. OpenAI, released 2026-03-17 alongside GPT-5.4 + nano. Efficiency/high-throughput variant of GPT-5.4 (anchor); brings core 5.4 capabilities at lower latency + cost. Independent sources: Artificial Analysis Intelligence Index = 49 (xhigh reasoning) / ~37.7 (non-reasoning) coding index ~37.5. Pricing $0.75/$4.50 per M (vs 5.4 Standard $2.50/$15). Speed ~150 t/s first-party (well above ~64 median for tier); Azure to 184 t/s. 400K-1M context (provider-dependent), 128K max output. Vision + text in. Knowledge cutoff Aug 2025. DIMENSION RATIONALE: scaled DOWN from GPT-5.4 across reasoning/coding/scientific (mini tier lags frontier per AA + Design-for-Online 'reasoning + coding lag top-tier'). Bulk 15 (vs 5.4's 18). Context Fidelity 16 (inherits large context). Speed 16 (genuinely fast, 150 t/s). Cost 15 (cheap but output $4.50/M is mid for mini-tier). Multimodal 14 (vision retained). Safety 15 (OpenAI alignment stack). Free-tier ChatGPT users get limited access. Anchored against GPT-5.4.