AI BENCHY Compare

OpenAI: GPT-5.4 vs StepFun: Step 3.7 Flash

Summary

The average score is effectively tied at 8.5 vs 8.5. Step 3.7 Flash (medium) has the lower benchmark cost at $0.376 vs $1.210. Step 3.7 Flash (medium) is faster at 20.35s vs 22.35s, with pass rates of 76.2% vs 73.0%.

Recommended modelStep 3.7 Flash (medium)It has the best score here (8.5), while costing about 3.2x less than GPT-5.4 (medium).

Last updated at: 2026-07-14

Metric	GPT-5.4 GPT-5.4 medium Release: 2026-03-05	Step 3.7 Flash Step 3.7 Flash medium Release: 2026-05-29

Metric	GPT-5.4 GPT-5.4 medium Release: 2026-03-05	Step 3.7 Flash Step 3.7 Flash medium Release: 2026-05-29
Score	8.5	8.5
Rank	#20	#23
Reliability	10.0	9.9
Consistency	8.6	9.3
Tests Correct
Attempt pass rate	76.2%	73.0%
Flaky tests	4	2
Total Runs	63	61
Cost per result	8.640	2.686
Total Cost	$1.210	$0.376
Input Price	$2.500 / 1M	$0.200 / 1M
Output Price	$15.000 / 1M	$1.150 / 1M
Total Input Tokens	34,108	39,981
Output Tokens	2,242	319,958
Reasoning Tokens	72,707	0
Response Time (avg)	22.35s	20.35s
Response Time (max)	100.41s	113.98s
Response Time (total)	469.29s	427.42s

Generation showcase

Hamster playing table tennis

Prompt: Create a detailed SVG illustration of a hamster playing table tennis.

#20 GPT-5.4

medium

Cost: $0.214
Time: 199.6s
Tokens: 14,349 tok

#23 Step 3.7 Flash

medium

Cost: $0.006
Time: 46.2s
Tokens: 4,466 tok

Top Models by Score

Score vs Total Cost

Response Time (avg)

Score vs Response Time (avg)

Total Output Tokens

Score vs Total Output Tokens

Category Breakdown

Category:

Anti-AI Tricks	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	8.3	10.0	75.0%	0		4.11s	606	240	1,511
Step 3.7 Flash	8.7	7.9	91.7%	1		9.65s	756	32,185	0

Coding	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	8.8	7.8	88.9%	1		44.36s	7,305	433	24,216
Step 3.7 Flash	8.8	7.8	88.9%	1		27.42s	7,437	44,797	0

Combined	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	10.0	10.0	100.0%	0		20.57s	11,019	301	3,543
Step 3.7 Flash	10.0	10.0	100.0%	0		9.06s	13,683	7,106	0

Data parsing and extraction	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	10.0	10.0	100.0%	0		5.32s	7,140	234	804
Step 3.7 Flash	10.0	10.0	100.0%	0		2.75s	7,398	3,020	0

Domain specific	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	5.3	7.2	44.4%	1		74.27s	619	61	34,748
Step 3.7 Flash	7.7	10.0	66.7%	0		48.27s	708	70,347	0

General Intelligence	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	4.7	3.1	33.3%	1		4.92s	477	145	321
Step 3.7 Flash	4.0	10.0	0.0%	0		6.85s	525	3,987	0

Instructions following	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	10.0	10.0	100.0%	0		3.11s	660	93	897
Step 3.7 Flash	9.8	10.0	100.0%	0		1.83s	735	2,166	0

Puzzle Solving	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	8.2	7.2	88.9%	1		9.14s	642	441	3,815
Step 3.7 Flash	5.7	9.9	33.3%	0		6.19s	756	15,071	0

Tool Calling	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	10.0	10.0	100.0%	0		13.28s	5,445	264	1,031
Step 3.7 Flash	10.0	10.0	100.0%	0		4.16s	7,746	2,115	0

Trivia	Score	Consistency	Attempt pass rate	Flaky tests	Tests Correct	Response Time (avg)	Input Tokens	Output Tokens	Reasoning Tokens
GPT-5.4	3.0	10.0	0.0%	0		13.95s	195	30	1,821
Step 3.7 Flash	3.0	10.0	0.0%	0		113.98s	237	139,164	0

Quick Compare

Switch Comparison Pair

GPT-5.2 ChatnonevsStep 3.7 Flashmedium DeepSeek V4 FlashhighvsStep 3.7 Flashmedium DeepSeek V4 FlashhighvsGPT-5.4medium GPT-5.6 TerrahighvsStep 3.7 Flashmedium Step 3.7 FlashmediumvsGLM 5.2high GPT-5.4mediumvsGLM 5.2high KAT-Coder-Pro V2.5lowvsStep 3.7 Flashmedium KAT-Coder-Pro V2.5lowvsGPT-5.4medium Step 3.7 FlashmediumvsGrok 4.5low GPT-5.4mediumvsGrok 4.5low Gemini 3.5 FlashlowvsGPT-5.4medium Claude Opus 4.8lowvsStep 3.7 Flashmedium