Compare Models
Select models (max 7)
Claude Opus 4 (Nonthinking)GPT-6.2 Astra
| Vals IndexGDP-weighted benchmark | N/A | 67.66%±0.94 |
| Legal Research BenchAgentic US legal research | N/A | 42.79%±3.44 |
| Finance Agent v2.1Core financial analyst tasks | N/A | 60.75%±2.12 |
| Tax Agent BenchAgentic US corporate tax research | N/A | 21.84%±2.79 |
| Terminal-Bench-ScienceExpert-authored scientific research workflows | N/A | 71.43%±5.44 |
| Terminal-Bench 4.0Frontier-difficulty terminal tasks | N/A | 63.13%±1.34 |
| Vibe Code Bench v1.1Building web apps from scratch | N/A | 89.90%±2.04 |
Benchmarks
Vals Index *
Claude Opus 4 (Nonthinking)
N/AGPT-6.2 Astra
0.00%± 0.94
(41/41)Legal Research Bench *
Claude Opus 4 (Nonthinking)
N/AGPT-6.2 Astra
0.00%± 3.44
(98/98)Finance Agent v2.1 *
Claude Opus 4 (Nonthinking)
N/AGPT-6.2 Astra
0.00%± 2.12
(44/44)Tax Agent Bench *
Claude Opus 4 (Nonthinking)
N/AGPT-6.2 Astra
0.00%± 2.79
(89/89)Terminal-Bench-Science
Claude Opus 4 (Nonthinking)
N/AGPT-6.2 Astra
0.00%± 5.44
(47/47)Terminal-Bench 4.0
Claude Opus 4 (Nonthinking)
N/AGPT-6.2 Astra
0.00%± 1.34
(68/68)Vibe Code Bench v1.1 *
Claude Opus 4 (Nonthinking)
N/AGPT-6.2 Astra
0.00%± 2.04
(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
83.07%
vs.29.52%
Finance
65.25%
vs.52.87%
Science
N/A
vs.68.44%
Academic
77.06%
vs.N/A
Coding
62.63%
vs.69.39%
Cost Analysis
Per task and per token model pricing.
Vals Index · USD
N/A
$10.68
Cost / TestVals Index
N/A
vs.$10.68
Input Cost/ 1M Tokens
$15.00
vs.$8.00
Input Cache Write/ 1M Tokens
$18.75
vs.$10.00
Input Cache Read/ 1M Tokens
$1.50
vs.$0.40
Output Cost/ 1M Tokens
$75.00
vs.$40.00
Model Metadata
Basic information about each model.
| Model provider | ||
| Latency | N/A | 3136.31s |
| Cost (In/Out) | $15 / $75 | $8 / $40 |
| Context Window | 200k | 1M |
| Max Output Token | 32k | 128k |
| Input Modality |