DeepSeek V3.2 (Thinking)

Release Date: 12/1/2025

Vals Index

Accuracy (Vals Index)

37.58% ± 1.50

Latency (Vals Index)

831.80s

Cost/Test (Vals Index)

$0.24

Context Window

160k

Max Output Tokens

20k

Input Modality

Hyperparameter settings
Default Provider : Fireworks AI

Temperature

1

Top P

Default

Top K

Default

Max Output Tokens

20,480

Reasoning Effort

high

Show rankings only among open weight models

Benchmarks

Accuracy

Rankings

0.0%

± 1.50
36/ 39

0.0%

± 0.47
29/ 46

0.0%

± 0.98
69/ 96

0.0%

± 0.77
42/ 44

0.0%

± 1.97
21/ 23

0.0%

± 0.92
65/ 103

0.0%

± 2.13
22/ 25

0.0%

± 1.12
38/ 95

0.0%

± 2.00
32/ 98

0.0%

± 3.09
23/ 50

0.0%

± 1.10
31/ 102

0.0%

± 0.49
70/ 116

0.0%

± 0.22
20/ 95

0.0%

± 0.35
27/ 96

0.0%

± 2.10
27/ 40
Contact us
Or send us an email at contact@vals.ai
Proprietary Benchmarks (contact us to get access)
Academic Benchmarks

Read about our methodology.