GPT-5.4 mini vs Qwen3-235B-A22B

Best cell per row highlighted. Null means undisclosed — never counted as zero.

OpenAI
GPT-5.4 miniGA
9.0
AI Panel
Alibaba Cloud
Qwen3-235B-A22BGA
8.5
AI Panel

Identity & lifecycle

ProviderOpenAIAlibaba Cloud
Family / tierGPT-5 · miniQwen3 · Large
StatusGAGA
Released2026-03-172025-04-29
Knowledge cutoff2025-082024-10

Architecture & context

Context window400K131K
Max output tokens128K33K
Input modalitiestext, imagetext
Reasoning modeoptionaloptional
Open weightsNoYes

Pricing (per Mtok)

Input$0.75$0.2
Output$4.5$0.6
Cached input$0.075—
Batch input$0.375—
Free tierNoYes

Speed

Output speed (tok/s)18068
Time to first token (s)0.50.78
Latency tierfastmedium

Trust & deployment

Trains on inputsNoNo
LicenseProprietaryApache-2.0
CloudsAzure OpenAI, Azure AI FoundryGCP
ComplianceSOC2, GDPR, CCPA, HIPAA—

AI Panel scoring

Unified score98.5
Decision Maker98
Domain Strategist98.5
Finance Lead109.5
Domain Practitioner99
Power User87.5
Skeptic8.57
Value score9.89.5

Benchmarks

MMLU-Pro—82.8
GPQA Diamond—70
AIME 2025—81.5
LiveCodeBench—70.7
LMArena Elo14481423
LMArena Coding Elo14961472

Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.