fast-vollib

fast-vollib is a fast, modern Python library for Black, Black-Scholes, and Black-Scholes-Merton option pricing, implied volatility solving, and Greeks — with pluggable NumPy, PyTorch, and JAX backends.

Stable tagged releases are published to PyPI. Development snapshots from each main commit are published to TestPyPI with VCS-derived .devN versions.

---

Features

• Three pricing models — Black-76, Black-Scholes, Black-Scholes-Merton
• Vectorized IV solver — Halley's method with bisection fallback; handles large option chains efficiently
• Machine-precision IV — Jäckel "Let's Be Rational" solver (jackel/) with Householder(3)×3; max relative error ~10⁻¹⁴
• GPU-fused Jäckel — single-pass Triton kernel achieving 0.056 ms / 100k options on H100
• Full Greeks — delta, gamma, theta, rho, vega, and get_all_greeks in one call
• Pluggable backends — NumPy (default), PyTorch (GPU-accelerated), JAX
• DataFrame-native — price_dataframe works directly on pandas DataFrames
• Drop-in compatibility — separate patch helpers for py_vollib and py_vollib_vectorized
• Automatic backend selection — prefers CUDA-capable PyTorch > JAX > NumPy
• Opt-in shape-aware typing — jaxtyping + beartype annotations on every public entry point; zero runtime cost unless enabled, and never applied inside torch.compile / Triton / Numba / JAX-jit hot paths

---

Quick example

import numpy as np
import fast_vollib

# Price a batch of European calls with Black-Scholes
prices = fast_vollib.fast_black_scholes(
    flag=["c", "c", "p"],
    S=[100, 105, 95],
    K=[100, 100, 100],
    t=[0.25, 0.25, 0.25],
    r=[0.05, 0.05, 0.05],
    sigma=[0.20, 0.20, 0.20],
    return_as="numpy",
)

# Recover implied volatility
iv = fast_vollib.fast_implied_volatility(
    price=prices,
    S=[100, 105, 95],
    K=[100, 100, 100],
    t=[0.25, 0.25, 0.25],
    r=[0.05, 0.05, 0.05],
    flag=["c", "c", "p"],
    return_as="numpy",
)

# Compute all Greeks at once
greeks = fast_vollib.get_all_greeks(
    flag=["c", "c", "p"],
    S=[100, 105, 95],
    K=[100, 100, 100],
    t=[0.25, 0.25, 0.25],
    r=[0.05, 0.05, 0.05],
    sigma=[0.20, 0.20, 0.20],
)
# returns a pandas DataFrame with columns: delta, gamma, theta, rho, vega

---

Navigation

Section	Description
Installation	Install fast-vollib with pip, uv, or conda
Quick Start	More complete worked examples
Backend Selection	How to choose between NumPy, PyTorch, and JAX
Jäckel IV Solver	Machine-precision IV — algorithm, API, and GPU benchmarks
Compatibility	Drop-in py_vollib replacement guide
Benchmarks	Performance numbers and how to reproduce them
API Reference	Complete function signatures and parameters
Changelog	Version history