GPT-5.4 mini vs Claude Haiku 4.5

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

OpenAI
GPT-5.4 miniGA
9.0
AI Panel
Anthropic
Claude Haiku 4.5GA
8.4
AI Panel

Identity & lifecycle

ProviderOpenAIAnthropic
Family / tierGPT-5 · miniClaude 4 · Haiku
StatusGAGA
Released2026-03-172025-10-15
Knowledge cutoff2025-082025-02

Architecture & context

Context window400K200K
Max output tokens128K64K
Input modalitiestext, imagetext, image
Reasoning modeoptionaloptional
Open weightsNoNo

Pricing (per Mtok)

Input$0.75$1
Output$4.5$5
Cached input$0.075$0.1
Batch input$0.375$0.5
Free tierNoYes

Speed

Output speed (tok/s)180101.8
Time to first token (s)0.50.86
Latency tierfastfast

Trust & deployment

Trains on inputsNoNo
LicenseProprietaryProprietary
CloudsAzure OpenAI, Azure AI FoundryBedrock, Vertex AI, Azure AI Foundry
ComplianceSOC2, GDPR, CCPA, HIPAASOC2, HIPAA, GDPR, ISO27001

AI Panel scoring

Unified score98.4
Decision Maker98.5
Domain Strategist98.5
Finance Lead1010
Domain Practitioner98.5
Power User88
Skeptic8.57.5
Value score9.89.8

Benchmarks

MMLU-Pro—76
GPQA Diamond—67.2
SWE-bench Verified—73.3
Terminal-Bench—41.75
LMArena Elo14481415
LMArena Coding Elo14961482

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