InFine.QuantFinance
Pricing, risk and calibration, with code you can read and run. Developed by in finē to share quantitative methods and explore their execution on CPU and GPU.
using InFine.QuantFinance.ClosedForm;
var call = BlackScholes.Greeks(
OptionType.Call,
spot: 100, strike: 100,
maturity: 1.0, rate: 0.025,
dividendYield: 0.01,
volatility: 0.20);
Console.WriteLine(call.Price);
// 8.586530 (rounded)Illustrative inputs. Expected output from the repository example.
Open the exampleWhy we are opening the code
This project started as a training resource for our quants and quant developers: a way to connect financial models, their implementation and the realities of computation.
Previously shared through a private repository with our clients, it is now used in several financial institutions. We have cleaned up the code and released it under the MIT licence so that anyone can study it, use it and contribute.
Start with a working example
The repository includes runnable .NET examples with synthetic data and expected outputs. Follow a calculation, inspect its assumptions and try it on your machine.
Pricing · Greeks · Implied volatility
Black-Scholes-Merton prices, sensitivities, implied volatility and Heston pricing. A first route from model inputs to numerical results.
Open the example — Price an option and its GreeksMonte Carlo · VaR / ES · CPU / GPU
Full-revaluation Monte Carlo VaR and Expected Shortfall, with risk decomposition. Compare CPU and GPU results when the GPU setup is available.
Open the example — Explore portfolio riskSVI · Calibration · Arbitrage checks
Calibrate SVI smiles, inspect butterfly and calendar arbitrage checks, and price from an interpolated surface.
Open the example — Fit a volatility surfaceAlso in the library: yield curves, Hull-White, portfolio construction, backtesting and an Excel add-in.
Browse all examplesMeasured, with the trade-offs
The compute kernels run on .NET. Hybridizer can compile those same kernels to CUDA for NVIDIA GPUs. The repository checks GPU results against their CPU counterparts.
| Workload | CPU · all cores | GPU · end to end |
|---|---|---|
| Monte Carlo VaR20M scenarios × 100 options × 101 factors | 12.96 s | 5.73 s |
| Bootstrap backtest1M paths × 2,520 days × 10 assets | 5.27 s | 2.62 s |
| Heston COS2M options | 8.67 s | 3.88 s |
| Black-Scholes1M options · transfer-bound | 5.9 ms | 23.9 ms |
Ryzen 9 7845HX (24 threads), RTX 4070 Laptop (8 GB), Windows. .NET 8, BenchmarkDotNet 0.15.8, double precision. GPU timings include host-device transfers.
The GPU helps on these heavy workloads. On the smaller Black-Scholes workload, data transfers make the CPU faster. Your workload and hardware determine the result.
Read the measurements and methodologyUse it, inspect it, contribute
Start on CPU with the .NET 8 SDK. Follow the GPU guide for the additional NVIDIA and Hybridizer setup, or explore the Excel examples.
Install the pricing package
dotnet add package InFine.QuantFinance.ClosedFormThe library is released under MIT. Hybridizer is a separate product with its own terms; its publisher asks you to contact them for commercial use. Third-party dependencies retain their respective licences.
Licence and third-party noticesThe project is pre-1.0 and its public API may change between minor versions. Multi-curve rates, credit and FX are outside the current scope. Review the documented limitations and validate the models for your intended use.
Status and known limitationsThe people behind the code
Adapting a model, integrating a calculation or understanding a performance bottleneck: these are the kinds of problems our consultants work on with financial institutions.