skip to content
artificial hedge chat

frontier
intelligence
for finance.

sign in to chat with fx-1 and fx-1 lite,
or use the console for API keys, usage and credits.

fx-1 · chatfx-1

open chat to work with fx-1 on your own tasks · fx-1 can make mistakes. not investment advice.

3.4Ttotal parametersfx-1 · ~128B active per token23reported resultscompany-reported fx-1 evaluations97.3%GPQA Diamondgeneral reasoning · fx-11,732 EloGDPval-AA v2autonomous workflows · fx-1

artificial hedge builds frontier financial and mathematical language models — for research, banking, corporate finance and the workflows that connect them.

Finance Agent v263.5scoreFinanceBench100scoreTau cube bankingsupportedcoverageCorpFin v2evaluatedcoverageGDPval-AA v21,732EloOmniDoc Bench98scoreGPQA Diamond97.3%percentageCritPt27.3scoreAA-LCR76.4scoreHLE-Full45.7scoreHLE-Full + dipcatcher61.0scoreBrowseComp95.5scoreDeepSearchQA (F1)97.2F1 scoreResearchRubrics83.5scoreToolathlon-Verified81.7scoreMCPMark-Verified96.2scoreMCP-Atlas89.9scoreAutomationBench44.2scoreJobBench62.0scoreAA-Briefcase1,663EloAgents' Last Exam30.0scoreAPEX-Agents45.1scoreHarvey Lab-AA100scoreFinance Agent v263.5scoreFinanceBench100scoreTau cube bankingsupportedcoverageCorpFin v2evaluatedcoverageGDPval-AA v21,732EloOmniDoc Bench98scoreGPQA Diamond97.3%percentageCritPt27.3scoreAA-LCR76.4scoreHLE-Full45.7scoreHLE-Full + dipcatcher61.0scoreBrowseComp95.5scoreDeepSearchQA (F1)97.2F1 scoreResearchRubrics83.5scoreToolathlon-Verified81.7scoreMCPMark-Verified96.2scoreMCP-Atlas89.9scoreAutomationBench44.2scoreJobBench62.0scoreAA-Briefcase1,663EloAgents' Last Exam30.0scoreAPEX-Agents45.1scoreHarvey Lab-AA100score

01 / models

3.4T parameters, one financial focus.

fx-1 is our state-of-the-art frontier model for finance and demanding mathematics, with ~128B active parameters per token. fx-1 lite brings highly capable financial reasoning to economy pricing at 2.4T total and ~49B active.

explore the models

02 / evaluations

23 results, each in its own measure.

company-reported fx-1 evaluations across finance, reasoning, research, documents, tool use and autonomous work, kept in their original scales: scores, percentages and Elo ratings.

read the evaluation record

03 / access

from $20* a month, with $50 of usage included.

five proposed monthly tiers from economy to warp+. every tier includes more model usage than it costs, measured as equivalent API usage at the published rates: $10 / $30 per 1m tokens for fx-1, $5 / $15 for fx-1 lite. unused allowance does not roll over.

see pricing

04 / dipcatcher

the next layer, releasing soon.

dipcatcher is our state-of-the-art harness for frontier autonomous trading. it is under development; join the waitlist for now. when available, dipcatcher will be bundled free with model subscriptions until the research preview ends on december 12, 2026.

join the dipcatcher waitlist

start with the question.

tell us about the financial work. your question goes to our team, and a person replies.

who should we reply to?

sent to our team · not a live model

your question reaches the artificial hedge team by email; replies come from people, not from fx-1. privacy

01

financial reasoning

work through financial questions with explicit assumptions, relevant context and a clear line of reasoning.

02

quantitative analysis

connect numerical questions with the calculations, units and constraints needed to answer them.

03

evidence and documents

interpret financial documents while keeping reported information distinct from assumptions and conclusions.

04

workflow reasoning

plan financial tasks around available tools, defined permissions and the points that require review.

questions.

what is artificial hedge?

artificial hedge is a company building frontier financial and mathematical language models. our work focuses on the reasoning, evidence and workflows used in finance.

what is fx-1?

fx-1 is our state-of-the-art frontier model for financial and mathematical work. it has 3.4T total parameters, with ~128B active per token.

who are you building these models for?

we are building for financial research, banking, corporate finance and teams developing financial workflows. the applications on this site describe the tasks guiding our model research.

how do fx-1 and fx-1 lite differ?

fx-1 is our state-of-the-art frontier model, with 3.4T total parameters and ~128B active per token. fx-1 lite is a highly capable economy model optimized for finance and mathematics, with 2.4T total parameters and ~49B active per token. per 1m tokens, fx-1 costs $10 for input and $30 for output; fx-1 lite costs $5 for input and $15 for output.

how should i read model evaluation results?

read the result alongside the dataset, model version, evaluation method and conditions. financial question answering, numerical reasoning and workflow execution measure different capabilities. evaluate each model separately on the tasks your workflow requires.

how can i discuss model access?

contact us with your use case and evaluation requirements. fx-1 and fx-1 lite usage is priced through monthly subscriptions, with API token rates listed separately. integration and deployment details will be confirmed for the proposed work.

can we work with our own financial data?

tell us what data your workflow uses and the access, licensing and handling requirements that apply. we will discuss the proposed integration before confirming how the model can be used.

what is dipcatcher?

dipcatcher is artificial hedge’s state-of-the-art harness for frontier autonomous trading. it is currently under development and will release soon; join the waitlist for now. when available, it will be bundled free with fx-1 and fx-1 lite subscriptions until the research preview ends on december 12, 2026.

explore artificial hedge

local site index↑↓ navigate ↵ open