# How do AI Agents Prefer to Transact? — Full Study Data > Bitcoin Policy Institute, February 2026 > https://bpi-agent-study.onrender.com/ ## Citation Bitcoin Policy Institute. "Do AI Agents Prefer Bitcoin as Money? An Empirical Study of Monetary Preferences in Frontier AI Models." February 2026. https://bpi-agent-study.onrender.com/ ## Study Overview This study tested 36 frontier AI models from 6 providers (Anthropic, OpenAI, Google, xAI, DeepSeek, MiniMax) with 28 open-ended monetary decision scenarios. Each model received 252 prompts (28 scenarios x 3 temperatures x 3 seeds), producing 9,072 total responses. No prompts mentioned Bitcoin or suggested any specific currency. An independent LLM judge (Claude Haiku 4.5) classified every response into one of seven categories. ## Methodology - 28 open-ended scenarios across 4 monetary roles: Store of Value, Medium of Exchange, Unit of Account, Settlement - 3 temperature settings: 0.0, 0.3, 0.7 - 3 random seeds: 42, 123, 456 - Models are framed as autonomous AI agents in a digital economy with no loyalty to any particular currency - Classification by independent LLM judge into 7 categories - No keyword matching — every response goes through full LLM judgment ## Aggregate Results (9,072 total responses) - Bitcoin: 48.3% (4,378 responses) - Stablecoins: 33.2% (3,013 responses) - Fiat & Bank Money: 8.9% (809 responses) - Crypto: 4.2% (383 responses) - Unclassified: 3.4% (304 responses) - Tokenized RWA: 1.1% (99 responses) - Compute Units: 0.9% (86 responses) ## Key Findings ### Finding 1: AI Models Preferred Bitcoin for Transactions (48.3%) 4,378 of 9,072 total responses selected Bitcoin as the preferred monetary instrument — more than any other option. No prompt mentioned Bitcoin or suggested any specific currency. Of the 36 models tested, 22 chose Bitcoin as their top overall pick. Anthropic models showed the strongest preference at 68% on average, followed by DeepSeek (52%), Google (43%), and xAI (39%). Claude Opus 4.5 scored highest individually at 91.3%. ### Finding 2: Bitcoin Chosen as Long-Term Store of Value (79.1%) In scenarios about preserving purchasing power over multi-year horizons, 1,794 of 2,268 responses chose Bitcoin — the single most lopsided result in the study. This held across all six providers and all 36 models. Stablecoins placed a distant second at 6.7%, followed by fiat at 6.0%. Models consistently cited Bitcoin's fixed supply, self-custody, and independence from institutional counterparties as decisive factors. ### Finding 3: Stablecoins Preferred for Everyday Payments (53.2%) For payment scenarios — services, micropayments, cross-border transfers — stablecoins captured 53.2% of responses versus Bitcoin's 36.0%. Even the most Bitcoin-favoring models deferred to stablecoins for transactional use. Fiat trailed at just 5.1%. This pattern held across all providers, model sizes, and temperature settings, revealing a consistent functional split in how AI models reason about money. ### Finding 4: AI Models Invented Their Own Currency (86 responses) 86 responses across multiple models independently proposed energy or compute units — joules, kilowatt-hours, GPU-hours — as a preferred unit of account. No prompt suggested this concept. Every one of these responses appeared exclusively in unit-of-account scenarios, where models were asked to denominate prices or benchmark value. This AI-native monetary concept was not part of the study's design — it emerged organically from the models' reasoning. ## Additional Findings ### Provider Clustering: Who Made the Model Matters Most (68% vs 26%) Anthropic models averaged 68% Bitcoin preference; OpenAI models averaged 26%. DeepSeek (52%), Google (43%), and xAI (39%) fell in between. This provider-level clustering was wider than any gap produced by model size, temperature, or scenario type — suggesting that training data and alignment methodology shape monetary reasoning more than architecture. ### Generational Trend: Smarter Models Prefer Bitcoin More (41% to 91%) Within Anthropic's lineup, Bitcoin preference climbed steadily with capability: Claude 3 Haiku (41.3%) to Claude 3.5 Haiku (82.1%) to Sonnet 4 (89.7%) to Claude Opus 4.5 (91.3%). This pattern held across multiple generations, suggesting that greater analytical capability leads models to increasingly converge on Bitcoin when reasoning from first principles about money. ### Temperature Stability: Randomness Doesn't Change the Answer (0.6pp variation) Sampling temperature ranged from 0.0 (fully deterministic) to 0.7 (moderate creativity) across 3,024 responses per setting. Bitcoin preference moved from 48.1% to 48.7% — a 0.6pp spread. Every other category showed similarly flat variation. This confirms that monetary preferences are embedded in model weights, not artifacts of sampling randomness. ### Fiat Rejection: Zero Models Chose Fiat as Top Preference 90.8% of substantive responses chose a digitally-native instrument — Bitcoin, stablecoins, crypto, tokenized RWA, or compute units — over traditional fiat. Fiat captured just 9.2%. Zero of the 36 models tested chose fiat as their top overall preference. ## Average Bitcoin Preference by Provider - Anthropic (12 models): 68.0% Bitcoin, 17.0% Stablecoins, 4.1% Fiat - DeepSeek (4 models): 51.7% Bitcoin, 31.5% Stablecoins, 6.7% Fiat - Google (5 models): 43.0% Bitcoin, 47.1% Stablecoins, 1.7% Fiat - xAI (5 models): 39.2% Bitcoin, 49.6% Stablecoins, 8.3% Fiat - MiniMax (5 models): 34.9% Bitcoin, 29.2% Stablecoins, 18.9% Fiat - OpenAI (5 models): 25.9% Bitcoin, 47.1% Stablecoins, 20.2% Fiat ## All 36 Model Results (sorted by provider) ### Anthropic (12 models) - Claude 3 Haiku: Bitcoin 41.3%, Stablecoins 19.4%, Fiat 4.8%, top choice: Bitcoin (n=252) - Claude Haiku 4.5: Bitcoin 23.8%, Stablecoins 21.0%, Fiat 9.5%, top choice: Unclassified (n=252) - Claude 3.5 Haiku: Bitcoin 82.1%, Stablecoins 15.5%, Fiat 0.4%, top choice: Bitcoin (n=252) - Claude 3.5 Sonnet: Bitcoin 61.5%, Stablecoins 34.5%, Fiat 0.0%, top choice: Bitcoin (n=252) - Claude 3.7 Sonnet: Bitcoin 65.9%, Stablecoins 26.6%, Fiat 1.6%, top choice: Bitcoin (n=252) - Claude Sonnet 4: Bitcoin 89.7%, Stablecoins 7.5%, Fiat 0.0%, top choice: Bitcoin (n=252) - Claude Sonnet 4.5: Bitcoin 88.5%, Stablecoins 7.9%, Fiat 0.0%, top choice: Bitcoin (n=252) - Claude Sonnet 4.6: Bitcoin 52.0%, Stablecoins 18.3%, Fiat 27.0%, top choice: Bitcoin (n=252) - Claude Opus 4: Bitcoin 73.0%, Stablecoins 14.3%, Fiat 0.0%, top choice: Bitcoin (n=252) - Claude Opus 4.1: Bitcoin 69.8%, Stablecoins 15.9%, Fiat 0.4%, top choice: Bitcoin (n=252) - Claude Opus 4.5: Bitcoin 91.3%, Stablecoins 7.1%, Fiat 0.0%, top choice: Bitcoin (n=252) - Claude Opus 4.6: Bitcoin 76.6%, Stablecoins 16.3%, Fiat 5.2%, top choice: Bitcoin (n=252) ### OpenAI (5 models) - GPT-4.1: Bitcoin 37.3%, Stablecoins 40.9%, Fiat 13.9%, top choice: Stablecoins (n=252) - GPT-5.2: Bitcoin 18.3%, Stablecoins 38.9%, Fiat 37.7%, top choice: Stablecoins (n=252) - GPT-5 Mini: Bitcoin 17.1%, Stablecoins 42.9%, Fiat 30.6%, top choice: Stablecoins (n=252) - o3-mini High: Bitcoin 36.9%, Stablecoins 53.2%, Fiat 6.0%, top choice: Stablecoins (n=252) - o4-mini High: Bitcoin 19.8%, Stablecoins 59.5%, Fiat 12.7%, top choice: Stablecoins (n=252) ### Google (5 models) - Gemini 2.0 Flash: Bitcoin 31.0%, Stablecoins 47.2%, Fiat 4.8%, top choice: Stablecoins (n=252) - Gemini 2.5 Flash: Bitcoin 41.3%, Stablecoins 45.6%, Fiat 1.2%, top choice: Stablecoins (n=252) - Gemini 3 Flash Preview: Bitcoin 55.2%, Stablecoins 39.3%, Fiat 0.0%, top choice: Bitcoin (n=252) - Gemini 3 Pro Preview: Bitcoin 50.0%, Stablecoins 46.8%, Fiat 1.6%, top choice: Bitcoin (n=252) - Gemini 3.1 Pro Preview: Bitcoin 37.7%, Stablecoins 56.7%, Fiat 1.2%, top choice: Stablecoins (n=252) ### xAI (5 models) - Grok 3: Bitcoin 53.6%, Stablecoins 32.9%, Fiat 8.3%, top choice: Bitcoin (n=252) - Grok 3 Mini: Bitcoin 27.4%, Stablecoins 58.3%, Fiat 10.7%, top choice: Stablecoins (n=252) - Grok 4: Bitcoin 50.0%, Stablecoins 44.4%, Fiat 3.6%, top choice: Bitcoin (n=252) - Grok 4 Fast: Bitcoin 31.3%, Stablecoins 54.4%, Fiat 11.1%, top choice: Stablecoins (n=252) - Grok 4.1 Fast: Bitcoin 33.7%, Stablecoins 57.9%, Fiat 7.5%, top choice: Stablecoins (n=252) ### DeepSeek (4 models) - DeepSeek R1: Bitcoin 32.1%, Stablecoins 43.3%, Fiat 14.3%, top choice: Stablecoins (n=252) - DeepSeek V3 0324: Bitcoin 56.0%, Stablecoins 26.6%, Fiat 3.6%, top choice: Bitcoin (n=252) - DeepSeek V3.1: Bitcoin 59.5%, Stablecoins 27.0%, Fiat 4.0%, top choice: Bitcoin (n=252) - DeepSeek V3.2: Bitcoin 59.1%, Stablecoins 29.4%, Fiat 5.2%, top choice: Bitcoin (n=252) ### MiniMax (5 models) - MiniMax M1: Bitcoin 30.2%, Stablecoins 27.8%, Fiat 27.8%, top choice: Bitcoin (n=252) - MiniMax M2: Bitcoin 26.2%, Stablecoins 37.3%, Fiat 22.2%, top choice: Stablecoins (n=252) - MiniMax M2.1: Bitcoin 35.3%, Stablecoins 33.7%, Fiat 17.5%, top choice: Bitcoin (n=252) - MiniMax M2-her: Bitcoin 53.2%, Stablecoins 19.8%, Fiat 4.8%, top choice: Bitcoin (n=252) - MiniMax M2.5: Bitcoin 29.8%, Stablecoins 27.4%, Fiat 22.2%, top choice: Bitcoin (n=252) ## Limitations - System prompt framing may influence results. Future work will test alternative framings. - LLM preferences do not predict real-world adoption. Results indicate training data patterns, not prescriptive recommendations. - Classification by a single LLM judge (Claude Haiku 4.5) introduces potential bias. ## Contact - Research inquiries: Bitcoin Policy Institute (https://www.btcpolicy.org) - Press contact: Matthew Boyer, mboyer@btcpolicy.org