Compare Models
Select models (max 5)
GPT-6 AstraClaude Haiku 5.5
 |  |  |
|---|---|---|
| Vals IndexGDP-weighted benchmark | 63.13%±1.19 | 54.31%±1.41 |
| Legal Research BenchAgentic US legal research | 39.42%±3.40 | 43.27%±3.44 |
| Finance Agent (v2)Core financial analyst tasks | 53.54%±2.08 | 54.14%±2.13 |
| Tax Agent BenchAgentic US corporate tax research | 20.72%±2.92 | 31.92%±3.35 |
| MedCodeMedical billing code support | 48.49%±2.13 | 51.54%±2.17 |
| Terminal-Bench ScienceExpert-authored scientific research workflows | 62.86%±5.82 | 10.00%±3.61 |
| Code MigrationRewriting programs in new languages | 67.74%±4.22 | 44.57%±4.23 |
| Terminal-Bench 4.0Frontier-difficulty terminal tasks | 59.60%±4.40 | 35.35%±2.20 |
| Vibe Code Bench v1.1Building web apps from scratch | 89.59%±2.17 | 90.44%±1.73 |
Benchmarks
Vals Index *
GPT-6 Astra
0.00%± 1.19
(45/45)Claude Haiku 5.5
0.00%± 1.41
(45/45)Legal Research Bench *
GPT-6 Astra
0.00%± 3.40
(75/75)Claude Haiku 5.5
0.00%± 3.44
(75/75)Finance Agent (v2) *
GPT-6 Astra
0.00%± 2.08
(76/76)Claude Haiku 5.5
0.00%± 2.13
(76/76)Tax Agent Bench *
GPT-6 Astra
0.00%± 2.92
(67/67)Claude Haiku 5.5
0.00%± 3.35
(67/67)MedCode *
GPT-6 Astra
0.00%± 2.13
(106/106)Claude Haiku 5.5
0.00%± 2.17
(106/106)Terminal-Bench Science
GPT-6 Astra
0.00%± 5.82
(39/39)Claude Haiku 5.5
0.00%± 3.61
(39/39)Code Migration *
GPT-6 Astra
0.00%± 4.22
(75/75)Claude Haiku 5.5
0.00%± 4.23
(75/75)Terminal-Bench 4.0
GPT-6 Astra
0.00%± 4.40
(45/45)Claude Haiku 5.5
0.00%± 2.20
(45/45)Vibe Code Bench v1.1 *
GPT-6 Astra
0.00%± 2.17
(110/110)Claude Haiku 5.5
0.00%± 1.73
(110/110)Overall performance
Performance on the Vals Index, a GDP-weighted aggregation of tasks across finance, coding, and law
•2026-10-07
Comparison by Industry
Model performance on different sections of the economy.
Legal
22.42%
vs.22.26%
Finance
48.65%
vs.48.36%
Healthcare
68.20%
vs.70.54%
Math
99.00%
vs.86.00%
Science
65.09%
vs.38.50%
Education
46.37%
vs.45.78%
Coding
58.35%
vs.40.09%
Cyber
48.97%
vs.40.60%
Social Mobility
N/A
vs.63.19%
Cost Analysis
Per task and per token model pricing.
Vals Index · USD
$18.46
$2.991
Cost / TestVals Index
$18.46
vs.$2.991
Input Cost/ 1M Tokens
$10.00
vs.$0.10
Input Cache Write/ 1M Tokens
$12.50
vs.$0.125
Input Cache Read/ 1M Tokens
$1.00
vs.$0.01
Output Cost/ 1M Tokens
$50.00
vs.$0.50
Model Metadata
Basic information about each model.
| | |
|---|---|---|
| Model provider | OpenAI | Anthropic |
| Latency | 1667.74s | 2836.53s |
| Cost (In/Out) | $10 / $50 | $0.1 / $0.5 |
| Context Window | 1M | 1M |
| Max Output Token | 128k | 128k |
| Input Modality |