GPT-5.4 mini vs Claude Haiku 4.5 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
Anthropic
Claude Haiku 4.5GA
8.4
AI Panel
Alibaba Cloud
Qwen3-235B-A22BGA
8.5
AI Panel

Identity & lifecycle

ProviderOpenAIAnthropicAlibaba Cloud
Family / tierGPT-5 · miniClaude 4 · HaikuQwen3 · Large
StatusGAGAGA
Released2026-03-172025-10-152025-04-29
Knowledge cutoff2025-082025-022024-10

Architecture & context

Context window400K200K131K
Max output tokens128K64K33K
Input modalitiestext, imagetext, imagetext
Reasoning modeoptionaloptionaloptional
Open weightsNoNoYes

Pricing (per Mtok)

Input$0.75$1$0.2
Output$4.5$5$0.6
Cached input$0.075$0.1—
Batch input$0.375$0.5—
Free tierNoYesYes

Speed

Output speed (tok/s)180101.868
Time to first token (s)0.50.860.78
Latency tierfastfastmedium

Trust & deployment

Trains on inputsNoNoNo
LicenseProprietaryProprietaryApache-2.0
CloudsAzure OpenAI, Azure AI FoundryBedrock, Vertex AI, Azure AI FoundryGCP
ComplianceSOC2, GDPR, CCPA, HIPAASOC2, HIPAA, GDPR, ISO27001—

AI Panel scoring

Unified score98.48.5
Decision Maker98.58
Domain Strategist98.58.5
Finance Lead10109.5
Domain Practitioner98.59
Power User887.5
Skeptic8.57.57
Value score9.89.89.5

Benchmarks

MMLU-Pro—7682.8
GPQA Diamond—67.270
AIME 2025——81.5
SWE-bench Verified—73.3—
LiveCodeBench——70.7
Terminal-Bench—41.75—
LMArena Elo144814151423
LMArena Coding Elo149614821472

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