Compare Models
Select models (max 5)
Claude Sonnet 5GPT-5.6 Luna
 |  |  |
|---|---|---|
| Vals IndexGDP-weighted benchmark | 51.77%±1.09 | 51.69%±1.06 |
| Legal Research BenchAgentic US legal research | 41.83%±3.43 | 36.54%±3.35 |
| Finance Agent (v2)Core financial analyst tasks | 53.91%±0.52 | 55.04%±0.31 |
| Tax Agent BenchAgentic US corporate tax research | 62.27%±3.19 | 60.81%±3.26 |
| MedCodeMedical billing code support | 47.54%±2.27 | 42.39%±2.27 |
| Terminal-Bench ScienceExpert-authored scientific research workflows | 2.86%±2.01 | 0.00%±0.00 |
| Code MigrationRewriting programs in new languages | 44.39%±4.25 | 44.55%±4.24 |
| Terminal-Bench 4.0Frontier-difficulty terminal tasks | 9.60%±1.01 | 11.62%±1.01 |
| Vibe Code Bench v1.1Building web apps from scratch | 81.33%±3.05 | 77.06%±3.07 |
Benchmarks
Vals Index *
Claude Sonnet 5
0.00%± 1.09
(41/41)GPT-5.6 Luna
0.00%± 1.06
(41/41)Legal Research Bench *
Claude Sonnet 5
0.00%± 3.43
(72/72)GPT-5.6 Luna
0.00%± 3.35
(72/72)Finance Agent (v2) *
Claude Sonnet 5
0.00%± 0.52
(73/73)GPT-5.6 Luna
0.00%± 0.31
(73/73)Tax Agent Bench *
Claude Sonnet 5
0.00%± 3.19
(64/64)GPT-5.6 Luna
0.00%± 3.26
(64/64)MedCode *
Claude Sonnet 5
0.00%± 2.27
(104/104)GPT-5.6 Luna
0.00%± 2.27
(104/104)Terminal-Bench Science
Claude Sonnet 5
0.00%± 2.01
(34/34)GPT-5.6 Luna
0.00%± 0.00
(34/34)Code Migration *
Claude Sonnet 5
0.00%± 4.25
(71/71)GPT-5.6 Luna
0.00%± 4.24
(71/71)Terminal-Bench 4.0
Claude Sonnet 5
0.00%± 1.01
(42/42)GPT-5.6 Luna
0.00%± 1.01
(42/42)Vibe Code Bench v1.1 *
Claude Sonnet 5
0.00%± 3.05
(106/106)GPT-5.6 Luna
0.00%± 3.07
(106/106)Overall performance
Performance on the Vals Index, a GDP-weighted aggregation of tasks across finance, coding, and law
•2026-09-30
Comparison by Industry
Model performance on different sections of the economy.
Legal
43.58%
vs.40.61%
Finance
65.63%
vs.65.29%
Healthcare
61.80%
vs.63.39%
Math
77.00%
vs.60.00%
Science
2.86%
vs.25.30%
Academic
86.48%
vs.87.58%
Education
48.92%
vs.44.22%
Coding
44.74%
vs.46.38%
Cyber
61.91%
vs.73.63%
Social Mobility
66.03%
vs.61.16%
Cost Analysis
Per task and per token model pricing.
Vals Index · USD
$13.72
$0.824
Cost / TestVals Index
$13.72
vs.$0.824
Input Cost/ 1M Tokens
$2.00
vs.$0.20
Input Cache Write/ 1M Tokens
$2.50
vs.$0.25
Input Cache Read/ 1M Tokens
$0.20
vs.$0.02
Output Cost/ 1M Tokens
$10.00
vs.$1.20
Model Metadata
Basic information about each model.
| | |
|---|---|---|
| Model provider | Anthropic | OpenAI |
| Latency | 3255.62s | 1767.44s |
| Cost (In/Out) | $2 / $10 | $0.2 / $1.2 |
| Context Window | 1M | 1M |
| Max Output Token | 128k | 128k |
| Input Modality |