Compare Models
Select models (max 5)
DeepSeek V4.1 FlashHy4 Preview
 |  |  |
|---|---|---|
| Vals IndexGDP-weighted benchmark | 51.32%±1.13 | 49.94%±1.15 |
| Legal Research BenchAgentic US legal research | 41.35%±3.42 | 45.19%±3.46 |
| Finance Agent (v2)Core financial analyst tasks | 53.48%±0.39 | 55.06%±0.31 |
| Tax Agent BenchAgentic US corporate tax research | 62.46%±3.15 | 63.71%±3.26 |
| MedCodeMedical billing code support | 41.17%±2.04 | 43.25%±2.13 |
| Terminal-Bench ScienceExpert-authored scientific research workflows | 0.00%±0.00 | 1.43%±1.43 |
| Code MigrationRewriting programs in new languages | 45.62%±4.29 | 47.43%±4.27 |
| Terminal-Bench 4.0Frontier-difficulty terminal tasks | 19.70%±1.75 | 8.08%±1.34 |
| Vibe Code Bench v1.1Building web apps from scratch | 84.74%±2.89 | 77.48%±4.04 |
Benchmarks
Vals Index *
DeepSeek V4.1 Flash
0.00%± 1.13
(43/43)Hy4 Preview
0.00%± 1.15
(43/43)Legal Research Bench *
DeepSeek V4.1 Flash
0.00%± 3.42
(73/73)Hy4 Preview
0.00%± 3.46
(73/73)Finance Agent (v2) *
DeepSeek V4.1 Flash
0.00%± 0.39
(74/74)Hy4 Preview
0.00%± 0.31
(74/74)Tax Agent Bench *
DeepSeek V4.1 Flash
0.00%± 3.15
(65/65)Hy4 Preview
0.00%± 3.26
(65/65)MedCode *
DeepSeek V4.1 Flash
0.00%± 2.04
(104/104)Hy4 Preview
0.00%± 2.13
(104/104)Terminal-Bench Science
DeepSeek V4.1 Flash
0.00%± 0.00
(38/38)Hy4 Preview
0.00%± 1.43
(38/38)Code Migration *
DeepSeek V4.1 Flash
0.00%± 4.29
(73/73)Hy4 Preview
0.00%± 4.27
(73/73)Terminal-Bench 4.0
DeepSeek V4.1 Flash
0.00%± 1.75
(43/43)Hy4 Preview
0.00%± 1.34
(43/43)Vibe Code Bench v1.1 *
DeepSeek V4.1 Flash
0.00%± 2.89
(108/108)Hy4 Preview
0.00%± 4.04
(108/108)Overall performance
Performance on the Vals Index, a GDP-weighted aggregation of tasks across finance, coding, and law
•2026-10-02
Comparison by Industry
Model performance on different sections of the economy.
Legal
43.76%
vs.46.04%
Finance
57.72%
vs.58.95%
Healthcare
63.34%
vs.63.42%
Math
54.00%
vs.75.00%
Science
29.65%
vs.35.34%
Education
47.88%
vs.N/A
Coding
39.57%
vs.38.47%
Cyber
37.23%
vs.33.82%
Social Mobility
64.28%
vs.68.61%
Cost Analysis
Per task and per token model pricing.
Vals Index · USD
$0.332
$1.439
Cost / TestVals Index
$0.332
vs.$1.439
Input Cost/ 1M Tokens
$0.30
vs.$0.834
Input Cache Read/ 1M Tokens
$0.006
vs.$0.042
Output Cost/ 1M Tokens
$1.20
vs.$2.501
Model Metadata
Basic information about each model.
| | |
|---|---|---|
| Model provider | DeepSeek | Openrouter |
| Latency | 1673.17s | 3656.53s |
| Cost (In/Out) | $0.3 / $1.2 | $0.834 / $2.501 |
| Context Window | 1M | 1M |
| Max Output Token | 384k | 64k |
| Input Modality |