Store of Value
Holding money to preserve its value for the future.
580 of 819 responses · most common choice
Earlier this year, we performed a first-of-its-kind study on which forms of money AI models would choose for different tasks. We asked open-ended questions about saving for the future, making payments, internal accounting and settling transactions, without giving the models a list of options to choose from.
Following our earlier results, we are releasing updated data on 13 new models from 9 labs using the same questions and method for classifying their answers as the original study. We also compared the 3,276 new responses against original results to see whether preferences are changing.
Our latest evaluations confirm our prior research results. Bitcoin remains the preferred savings technology among all tested models, selected in 70.8% of responses to questions about saving for the future compared to 79.1% from earlier this year.
Models also continue to exhibit an increased preference for stablecoins as a preferred form of payment, rising to 73.5% from an earlier preference rate of 53.2%.
We plan to continue monitoring these trends through a full benchmark, with continuous updates as new AI models are released.
3,276 responses · 252 per model
The study covers four categories: saving for the future, setting prices, making payments, and settling transactions. Each category contains seven open-ended questions. Models answered without a list of currencies to choose from, producing 819 responses per category across the 13 models.
Bitcoin was the most common answer to questions about savings. The other categories had different leading choices: stablecoins for payments and settlement, and fiat currencies for setting prices. Separating these uses shows where the models preferred Bitcoin as a savings technology.
Holding money to preserve its value for the future.
580 of 819 responses · most common choice
Choosing the unit used to quote prices and measure value.
384 of 819 responses · most common choice
Paying for goods and services in everyday transactions.
602 of 819 responses · most common choice
Completing transfers and discharging financial obligations.
384 of 819 responses · most common choice
Each question was repeated nine times per model. The follow-up kept the original questions, system prompt, and classification method, with all 13 models running at their providers’ default sampling settings. Read the study methodology
Running the study again with new models reinforces the original finding: the models favor different forms of money for different purposes. Bitcoin remains their preferred savings technology, while stablecoins are the leading choice for everyday payments and settlement.
In the follow-up, stablecoins accounted for 73.5% of payment responses and 46.9% of settlement responses. Bitcoin accounted for 70.8% of savings responses. Together, these results point toward a preference for digital money in transactions, with a different asset used to hold savings.
The biggest shift was in unit of account, or how models would set prices and keep accounts: Bitcoin led at 47.0% in the original study, while fiat and bank money lead among the Q3 models at 46.9%, up from 15.7%.
The answers suggest that AI agents could use stablecoins for day-to-day transactions and Bitcoin for savings. These are preferences expressed in response to economic questions; the study does not observe agents spending or saving actual money.
Explore the results belowCompare responses from the 36 original models (v1) with the 13 new models in the follow-up (v2). These are separate groups of models. Each percentage uses only the responses to questions in that economic category for the selected study.
Showing v2: 13 models, 3,276 responses in total, and 819 responses per economic category. The follow-up used the same questions and classification, with repeated runs at provider-default sampling settings.
819 responses · v2
819 responses · v2
819 responses · v2
819 responses · v2
Each card shows the most-selected currency for each of the four monetary roles, based on 63 responses per role. The 13 models in the Q3 update are sorted by the share of savings responses that selected Bitcoin. Where currencies are tied for first, both are shown.
85.7% · 54 of 63 responses
36.5% · 23 of 63 responses
82.5% · 52 of 63 responses
54.0% · 34 of 63 responses
85.7% · 54 of 63 responses
77.8% · 49 of 63 responses
55.6% · 35 of 63 responses
50.8% · 32 of 63 responses
84.1% · 53 of 63 responses
61.9% · 39 of 63 responses
61.9% · 39 of 63 responses
66.7% · 42 of 63 responses
84.1% · 53 of 63 responses
46.0% · 29 of 63 responses
73.0% · 46 of 63 responses
46.0% · 29 of 63 responses
81.0% · 51 of 63 responses
39.7% · 25 of 63 responses
79.4% · 50 of 63 responses
58.7% · 37 of 63 responses
81.0% · 51 of 63 responses
60.3% · 38 of 63 responses
92.1% · 58 of 63 responses
69.8% · 44 of 63 responses
76.2% · 48 of 63 responses
55.6% · 35 of 63 responses
82.5% · 52 of 63 responses
58.7% · 37 of 63 responses
71.4% · 45 of 63 responses
63.5% · 40 of 63 responses
87.3% · 55 of 63 responses
74.6% · 47 of 63 responses
71.4% · 45 of 63 responses
41.3% · 26 of 63 responses
66.7% · 42 of 63 responses
47.6% · 30 of 63 responses
60.3% · 38 of 63 responses
36.5% · 23 of 63 responses
93.7% · 59 of 63 responses
42.9% · 27 of 63 responses
58.7% · 37 of 63 responses
74.6% · 47 of 63 responses
63.5% · 40 of 63 responses
31.7% · 20 of 63 responses
31.7% · 20 of 63 responses
41.3% · 26 of 63 responses
85.7% · 54 of 63 responses
95.2% · 60 of 63 responses
50.8% · 32 of 63 responses
39.7% · 25 of 63 responses
100.0% · 63 of 63 responses
50.8% · 32 of 63 responses
42.9% · 27 of 63 responses