Benchmark suiteLatest source data: Jul 14, 2026Checked: July 16, 2026

Stepfun benchmarks.

A lab-level profile for Stepfun, covering tracked models, category leaders, open/proprietary mix, speed, value, source coverage, and portfolio depth.

All labs

Lab rank

#17

By available Arena average

Models tracked

6

2 open-weight models

Portfolio coverage

20%

Average matched benchmark signals

Fastest model

0.74s

Step 3.5 Flash

Lab profile

Portfolio shape, leaders, and model coverage.

The lab page now starts with the portfolio read, then drills into category leaders, value rankings, and model-level coverage rather than repeating source panels.

Portfolio read

Stepfun has 6 tracked benchmark profiles.

Lab #17

The portfolio combines 4 proprietary models and 2 open-weight models. Its average available Arena rank is 140.4, with a best category rank of #84. Average profile coverage is 20%, so model-level gaps remain visible in the cards below.

Best Arena model

Step 1o Turbo 202506

Best category rank #84

Highest coverage

Step 3.5 Flash

44% matched signals

Fastest response

Step 3.5 Flash

0.74s first token

Access mix

Open versus proprietary

Proprietary

4 / 67%

Open weights

2 / 33%

Value leader

Step 3.5 Flash

Category leaders

Best model by benchmark lens.

Each card shows the highest-ranked model from this lab in a composite benchmark suite, with the global rank and index left intact.

Value

Quality, price, latency, and speed

Step 3.5 Flash

#1681

value index

Coding

Code Arena, SWE-bench, and output speed

Step 3.5 Flash

#9324

coding index

Agents

Search, document, code, and latency

Step 3.5 Flash

#12100

agent proxy

Reasoning

Text, document, vision, and mode signals

Step 3.5 Flash

#11179

reasoning proxy

Context

Document, search, and text proxy

Step 3.5 Flash

#11779

context proxy

Stepfun value ranking table

Models from this lab with both quality and commercial price matches.

Top 1 of 1

#16
Step 3.5 Flash

step-3.5-flash

StepfunOpen weights
81value index
Price
$0.150blended / 1M tokens
Latency
0.74stime to first token
Speed
241 tok/soutput speed
Quality
79quality signal
ArenaArtificial Analysis

Model portfolio

Stepfun model coverage map.

Every tracked model keeps its best Arena rank, coverage score, commercial metrics, and suite placement visible in one scannable grid.

Step 1o Turbo 202506

step-1o-turbo-202506

Proprietary

Coverage

22%

Best Arena

#84

Vision

Value

Not listed

No price match

Coding

Not listed

No coding index

Latency

Not listed

Not listed

Not listed

Step 3

step-3

Open weights

Coverage

22%

Best Arena

#90

Vision

Value

Not listed

No price match

Coding

Not listed

No coding index

Latency

Not listed

Not listed

Not listed

Step 1o Vision 32K Highres

step-1o-vision-32k-highres

Proprietary

Coverage

11%

Best Arena

#98

Vision

Value

Not listed

No price match

Coding

Not listed

No coding index

Latency

Not listed

Not listed

Not listed

Step 1v 32K

step-1v-32k

Proprietary

Coverage

11%

Best Arena

#111

Vision

Value

Not listed

No price match

Coding

Not listed

No coding index

Latency

Not listed

Not listed

Not listed

Step 3.5 Flash

step-3.5-flash

Open weights

Coverage

44%

Best Arena

#134

Text

Value

#16

value index

Coding

#93

coding index

Latency

0.74s

241 tok/s

$0.150

Step 2 16K Exp 202412

step-2-16k-exp-202412

Proprietary

Coverage

11%

Best Arena

#204

Text

Value

Not listed

No price match

Coding

Not listed

No coding index

Latency

Not listed

Not listed

Not listed