StepFun: Step 3.5 Flash

Benchmarks gegenereerd uit Aibenchy-testsuites op : 19 feb 2026

Categorie-uitsplitsing

Categorie	Volledig geslaagde tests	Score	Consistentie	Slaagpercentage per poging	Instabiele tests	Redeneerscore	Kosten
Anti-AI Tricks	1/2	5.50	5.81	83.3%	1	10.00	$0.00000
Data parsing and extraction	1/2	5.00	10.00	50.0%	0	9.75	$0.00000
Domain specific	1/3	4.00	7.21	44.4%	1	8.44	$0.00000
Instructions following	2/2	10.00	10.00	100.0%	0	9.67	$0.00000
Puzzle Solving	0/3	2.00	4.96	33.3%	2	9.22	$0.00000

Vergelijk StepFun: Step 3.5 Flash met...

#17 · MiniMax

Redenering (medium)

Score: 5.08

Consistentie: 6.00

Slaagpercentage per poging: 61.1%

Instabiele tests: 6

Kosten per resultaat: 4.0276

Correcte tests: 5/12

Totale kosten: $0.20138

#19 · OpenAI

Zonder redenering

Score: 4.00

Consistentie: 9.98

Slaagpercentage per poging: 25.0%

Instabiele tests: 0

Kosten per resultaat: 0.0576

Correcte tests: 3/12

Totale kosten: $0.00173

#16 · Anthropic

Redenering (medium)

Score: 5.42

Consistentie: 8.60

Slaagpercentage per poging: 55.5%

Instabiele tests: 2

Kosten per resultaat: 12.8695

Correcte tests: 6/12

Totale kosten: $0.77217

Vergelijk StepFun: Step 3.5 Flash met...