stochastic-rs

Quantitative finance in Rust: stochastic process simulation, option pricing and calibration, volatility surfaces, fixed income and credit, risk, statistics, copulas and neural volatility surrogates. Generic over f32 / f64, SIMD on the CPU, CUDA and Metal back-ends where they pay off, and Python bindings via PyO3 that ship the same surface as the Rust crates.

Documentation

📖 stochastic.rust-dd.com is the reference; this README only gets you installed and running.

• Getting started — Rust and Python installation, first program
• Concepts — the traits (ProcessExt, DistributionExt, ModelPricer), seeding, feature flags, design philosophy
• Processes · Distributions · Copulas · Statistics · Quant · AI — the catalogues with selection guides
• GPU support — what runs on which device today, precision, an executed T4 run
• Python — the bindings, device=, NumPy interop
• Benchmarks · Migrating to v3 · Tutorials

What is inside

One workspace, one umbrella crate (stochastic-rs) that re-exports the sub-crates:

Installation

The Rust crates require Rust 1.89 or newer.

The workspace declares this minimum in rust-version. Cargo's edition-2024 resolver uses it when choosing compatible dependency versions, so a fresh resolution can select an older release of a dependency. The MSRV CI job checks the workspace's libraries, tests, examples, and benchmarks with default features disabled. Optional feature combinations are checked separately on current Rust.

[dependencies]
stochastic-rs = "3.0.0-rc.2"

Device back-ends and other optional parts are cargo features (cuda, metal, accelerate, ai, dual-stream-rng); the installation guide and the feature flags page list them with what each pulls in. Sub-crates can be depended on directly for lean builds.

pip install stochastic-rs

The wheels are CPU-only and carry the whole surface on Linux, macOS and Windows (linear algebra is pure Rust). A source build with a device back-end: maturin develop --release --features metal (or cuda) in a checkout.

Quickstart

use stochastic_rs::prelude::*;
use stochastic_rs::simd_rng::Unseeded;
use stochastic_rs::stochastic::diffusion::ou::Ou;
use stochastic_rs::quant::pricing::heston::HestonPricer;

fn main() {
    // Mean-reverting Ornstein-Uhlenbeck path: Ou::new(theta, mu, sigma, n, x0, t, seed)
    let ou = Ou::<f64>::new(2.0, 0.0, 1.0, 1_000, Some(0.0), Some(1.0), Unseeded);
    let path = ou.sample();
    println!("OU path points: {}", path.len());

    // Heston (1993) European option, closed form. The model holds only its own
    // parameters; the pricing query is passed to the call, so one model can
    // price a whole strike/maturity grid.
    // HestonPricer::new args: v0, rho, kappa, theta, sigma, lambda
    let pricer = HestonPricer::new(0.04, -0.5, 2.0, 0.04, 0.3, Some(0.0));
    // price_call/price_put args: s, k, r, q, tau
    let call = pricer.price_call(100.0, 100.0, 0.03, 0.0, 1.0);
    let put = pricer.price_put(100.0, 100.0, 0.03, 0.0, 1.0);
    println!("call={call:.4}, put={put:.4}");
}

import stochastic_rs as srs

# Mean-reverting OU path: PyOu(theta, mu, sigma, n, x0=None, t=None, seed=None, dtype=None, device=None)
path = srs.PyOu(2.0, 0.0, 1.0, 1000, x0=0.0, t=1.0, seed=42).sample()   # numpy.ndarray, shape (1000,)

# Heston European option, closed form
pricer = srs.HestonPricer(
    s=100, v0=0.04, k=100, r=0.03, kappa=2.0, theta=0.04, sigma=0.3,
    rho=-0.5, tau=1.0, q=0.0,
)
call, put = pricer.call_put()

A process samples on a device by re-typing it: Gbm::new(...).on::<Metal>() (Cuda, Accelerate), with handle.probe() to check the device first; from Python, device="metal" on the device-capable classes. The GPU support page has the support matrix, and notebooks/ a Colab notebook that runs the CUDA back-end on a free T4.

Benchmarks

Criterion suites live under benches/; the benchmarks page carries the numbers: the SIMD Normal sampler against rand_distr, fractional Gaussian noise on CPU, Accelerate, Metal and cuFFT, and the per-release speedups.

Citing

The concept DOI 10.5281/zenodo.21553307 always resolves to the latest release; CITATION.cff carries the version DOI of the current one.

Contributing

Bug reports, suggestions and pull requests are welcome on GitHub. The contributing page has the development rules; per-feature recipes (add-diffusion-process, adding-distribution, calibration-pattern, …) live under .claude/skills/.

License

MIT — see LICENSE.