Compare Models
Select models (max 7)
DeepSeek V4GPT 5.4 Mini
| Legal Research BenchAgentic US legal research | 23.08%±2.93 | 12.50%±2.30 |
| Tax Agent BenchAgentic US corporate tax research | 25.17%±3.06 | 8.81%±2.04 |
| MedCodeMedical billing code support | 40.45%±2.12 | N/A |
| Terminal-Bench 4.0Frontier-difficulty terminal tasks | 11.11%±1.34 | 2.52%±1.01 |
| Vibe Code Bench v1.1Building web apps from scratch | 49.93%±4.77 | 47.97%±5.61 |
Benchmarks
Legal Research Bench *
DeepSeek V4
0.00%± 2.93
(98/98)GPT 5.4 Mini
0.00%± 2.30
(98/98)Tax Agent Bench *
DeepSeek V4
0.00%± 3.06
(89/89)GPT 5.4 Mini
0.00%± 2.04
(89/89)MedCode *
DeepSeek V4
0.00%± 2.12
(107/107)GPT 5.4 Mini
N/ATerminal-Bench 4.0
DeepSeek V4
0.00%± 1.34
(68/68)GPT 5.4 Mini
0.00%± 1.01
(68/68)Vibe Code Bench v1.1 *
DeepSeek V4
0.00%± 4.77
(132/132)GPT 5.4 Mini
0.00%± 5.61
(132/132)Overall performance
Performance on the Vals Index, a GDP-weighted aggregation of tasks across finance, coding, legal, and tax
Comparison by Industry
Model performance on different sections of the economy.
Legal
51.70%
vs.12.50%
Finance
48.62%
vs.47.85%
Healthcare
57.80%
vs.N/A
Math
16.00%
vs.N/A
Academic
88.32%
vs.82.29%
Coding
46.20%
vs.40.99%
Social Mobility
62.92%
vs.N/A
Cost Analysis
Per task and per token model pricing.
Vals Index · USD
N/A
N/A
Cost / TestVals Index
N/A
vs.N/A
Input Cost/ 1M Tokens
$1.32
vs.$0.75
Input Cache Read/ 1M Tokens
$0.044
vs.$0.075
Output Cost/ 1M Tokens
$3.96
vs.$4.50
Model Metadata
Basic information about each model.
| Model provider | ||
| Latency | N/A | N/A |
| Cost (In/Out) | $1.32 / $3.96 | $0.75 / $4.5 |
| Context Window | 1M | 400k |
| Max Output Token | 384k | 128k |
| Input Modality |