Compare Models
Select models (max 5)
Claude Opus 5.5Claude Sonnet 5.5Claude Fable 5.1Claude Opus 5GPT-6 Astra
 |  | | | |  |
|---|---|---|---|---|---|
| Vals IndexGDP-weighted benchmark | 69.69%±0.94 | 69.22%±0.96 | 68.83%±1.08 | 67.21%±0.98 | 66.61%±1.09 |
| Legal Research BenchAgentic US legal research | 50.48%±3.48 | 48.08%±3.47 | 55.29%±3.46 | 55.29%±3.46 | 39.42%±3.40 |
| Finance Agent (v2)Core financial analyst tasks | 58.59%±0.17 | 58.10%±0.67 | 58.88%±2.06 | 58.63%±0.08 | 53.54%±2.08 |
| Tax Agent BenchAgentic US corporate tax research | 70.50%±3.15 | 73.39%±2.98 | 77.64%±2.83 | 75.06%±2.96 | 63.34%±3.15 |
| MedCodeMedical billing code support | 49.80%±2.27 | 52.92%±2.12 | 53.51%±2.17 | 63.57%±1.99 | 48.49%±2.13 |
| Terminal-Bench ScienceExpert-authored scientific research workflows | 48.57%±6.02 | 38.57%±5.86 | 34.29%±5.71 | 22.86%±5.05 | 65.71%±5.71 |
| Code MigrationRewriting programs in new languages | 66.65%±4.33 | 69.83%±4.26 | 54.61%±4.81 | 57.47%±4.37 | 67.74%±4.22 |
| Terminal-Bench 4.0Frontier-difficulty terminal tasks | 61.62%±1.01 | 53.03%±1.51 | 49.49%±3.07 | 45.45%±2.62 | 57.07%±3.07 |
| Vibe Code Bench v1.1Building web apps from scratch | 90.29%±1.53 | 92.39%±1.26 | 90.26%±1.57 | 88.40%±3.00 | 89.59%±2.17 |
Benchmarks
Vals Index *
Claude Opus 5.5
0.00%± 0.94
(66/66)Claude Sonnet 5.5
0.00%± 0.96
(66/66)Claude Fable 5.1
0.00%± 1.08
(66/66)Claude Opus 5
0.00%± 0.98
(66/66)GPT-6 Astra
0.00%± 1.09
(66/66)Legal Research Bench *
Claude Opus 5.5
0.00%± 3.48
(69/69)Claude Sonnet 5.5
0.00%± 3.47
(69/69)Claude Fable 5.1
0.00%± 3.46
(69/69)Claude Opus 5
0.00%± 3.46
(69/69)GPT-6 Astra
0.00%± 3.40
(69/69)Finance Agent (v2) *
Claude Opus 5.5
0.00%± 0.17
(70/70)Claude Sonnet 5.5
0.00%± 0.67
(70/70)Claude Fable 5.1
0.00%± 2.06
(70/70)Claude Opus 5
0.00%± 0.08
(70/70)GPT-6 Astra
0.00%± 2.08
(70/70)Tax Agent Bench *
Claude Opus 5.5
0.00%± 3.15
(61/61)Claude Sonnet 5.5
0.00%± 2.98
(61/61)Claude Fable 5.1
0.00%± 2.83
(61/61)Claude Opus 5
0.00%± 2.96
(61/61)GPT-6 Astra
0.00%± 3.15
(61/61)MedCode *
Claude Opus 5.5
0.00%± 2.27
(102/102)Claude Sonnet 5.5
0.00%± 2.12
(102/102)Claude Fable 5.1
0.00%± 2.17
(102/102)Claude Opus 5
0.00%± 1.99
(102/102)GPT-6 Astra
0.00%± 2.13
(102/102)Terminal-Bench Science
Claude Opus 5.5
0.00%± 6.02
(33/33)Claude Sonnet 5.5
0.00%± 5.86
(33/33)Claude Fable 5.1
0.00%± 5.71
(33/33)Claude Opus 5
0.00%± 5.05
(33/33)GPT-6 Astra
0.00%± 5.71
(33/33)Code Migration *
Claude Opus 5.5
0.00%± 4.33
(69/69)Claude Sonnet 5.5
0.00%± 4.26
(69/69)Claude Fable 5.1
0.00%± 4.81
(69/69)Claude Opus 5
0.00%± 4.37
(69/69)GPT-6 Astra
0.00%± 4.22
(69/69)Terminal-Bench 4.0
Claude Opus 5.5
0.00%± 1.01
(37/37)Claude Sonnet 5.5
0.00%± 1.51
(37/37)Claude Fable 5.1
0.00%± 3.07
(37/37)Claude Opus 5
0.00%± 2.62
(37/37)GPT-6 Astra
0.00%± 3.07
(37/37)Vibe Code Bench v1.1 *
Claude Opus 5.5
0.00%± 1.53
(104/104)Claude Sonnet 5.5
0.00%± 1.26
(104/104)Claude Fable 5.1
0.00%± 1.57
(104/104)Claude Opus 5
0.00%± 3.00
(104/104)GPT-6 Astra
0.00%± 2.17
(104/104)Overall performance
Performance on the Vals Index, a GDP-weighted aggregation of tasks across finance, coding, and law
•2026-09-27
Comparison by Industry
Model performance on different sections of the economy.
Legal
27.12%
vs.25.50%
vs.50.16%
vs.49.64%
vs.22.42%
Finance
68.34%
vs.69.07%
vs.71.99%
vs.70.89%
vs.62.86%
Healthcare
70.61%
vs.72.01%
vs.72.40%
vs.77.28%
vs.68.20%
Math
100.00%
vs.100.00%
vs.100.00%
vs.99.00%
vs.99.00%
Science
59.13%
vs.56.26%
vs.41.02%
vs.46.50%
vs.66.04%
Academic
N/A
vs.N/A
vs.92.15%
vs.91.64%
vs.N/A
Education
45.83%
vs.51.77%
vs.48.53%
vs.49.43%
vs.46.37%
Coding
60.41%
vs.74.58%
vs.59.03%
vs.61.52%
vs.57.92%
Beta
34.31%
vs.44.87%
vs.42.45%
vs.26.35%
vs.51.90%
Social Mobility
70.64%
vs.67.19%
vs.74.90%
vs.76.93%
vs.N/A
Cost Analysis
Per task and per token model pricing.
Vals Index · USD
$32.77
$20.80
$28.92
$18.81
$19.09
Cost / TestVals Index
$32.77
vs.$20.80
vs.$28.92
vs.$18.81
vs.$19.09
Input Cost/ 1M Tokens
$4.00
vs.$2.00
vs.$10.00
vs.$5.00
vs.$10.00
Input Cache Write/ 1M Tokens
$5.00
vs.$2.50
vs.$12.50
vs.$6.25
vs.$12.50
Input Cache Read/ 1M Tokens
$0.40
vs.$0.20
vs.$0.25
vs.$0.50
vs.$1.00
Output Cost/ 1M Tokens
$20.00
vs.$10.00
vs.$50.00
vs.$25.00
vs.$50.00
Model Metadata
Basic information about each model.
| | | | | |
|---|---|---|---|---|---|
| Model provider | Anthropic | Anthropic | Anthropic | Anthropic | OpenAI |
| Latency | 4320.73s | 4207.12s | 4576.31s | 3348.57s | 1510.53s |
| Cost (In/Out) | $4 / $20 | $2 / $10 | $10 / $50 | $5 / $25 | $10 / $50 |
| Context Window | 1M | 1M | 1M | 1M | 1M |
| Max Output Token | 128k | 128k | 128k | 128k | 128k |
| Input Modality |