Compare Models

Select models (max 7)
Step 5 PreviewQwen 4 Max
Benchmarks

Vals Index *

Step 5 Preview
0.00%± 1.16
(38/38)
Qwen 4 Max
0.00%± 1.01
(38/38)

Legal Research Bench *

Step 5 Preview
0.00%± 3.41
(95/95)
Qwen 4 Max
0.00%± 3.42
(95/95)

Finance Agent v2.1 *

Step 5 Preview
0.00%± 2.05
(41/41)
Qwen 4 Max
0.00%± 2.02
(41/41)

Tax Agent Bench *

Step 5 Preview
0.00%± 3.11
(86/86)
Qwen 4 Max
0.00%± 3.54
(86/86)

Terminal-Bench-Science

Step 5 Preview
0.00%± 2.01
(45/45)
Qwen 4 Max
N/A

Code Migration v1.1 *

Step 5 Preview
0.00%± 3.43
(30/30)
Qwen 4 Max
N/A

Terminal-Bench 4.0

Step 5 Preview
0.00%± 2.62
(65/65)
Qwen 4 Max
0.00%± 0.88
(65/65)

Vibe Code Bench v1.1 *

Step 5 Preview
0.00%± 4.27
(129/129)
Qwen 4 Max
0.00%± 2.27
(129/129)

Overall performance

Performance on the Vals Index, a GDP-weighted aggregation of tasks across finance, coding, legal, and tax

2026-11-15

Comparison by Industry

Model performance on different sections of the economy.

Legal
26.65%
31.30%
Finance
51.91%
59.29%
Math
42.00%
N/A
Science
29.58%
N/A
Coding
45.62%
67.57%
Cyber
0.38%
N/A
Social Mobility
58.12%
N/A

Cost Analysis

Per task and per token model pricing.

Vals Index · USD
$2.734
$9.33
Cost / TestVals Index
$2.734
$9.33
Input Cost/ 1M Tokens
$1.00
$1.50
Input Cache Read/ 1M Tokens
$0.05
$0.075
Output Cost/ 1M Tokens
$2.70
$4.50

Model Metadata

Basic information about each model.

Model provider
StepFunStepfun
AlibabaAlibaba
Latency3928.91s3879.45s
Cost (In/Out)$1 / $2.7$1.5 / $4.5
Context Window1M1M
Max Output Token1M262k
Input Modality