> ## Documentation Index
> Fetch the complete documentation index at: https://fhenix-docs-broken-sample-fixes.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Gas Costs and Benchmarks

> What FHE operations cost in gas, and where to find operation latency benchmarks

An FHE operation splits its cost in two. Your contract pays gas onchain to request the operation. The encrypted math itself runs offchain in the [compute pipeline](/deep-dive/cofhe-components/compute-pipeline) and costs no gas.

## Gas costs

A call like `FHE.add` records a task through the TaskManager and returns a handle for the result. The gas covers that request and the access-control bookkeeping, not the FHE computation.

We are measuring the gas of every FHE operation per encrypted type. This page gets the full table when the measurements land.

## Performance benchmarks

The compute pipeline executes operations with [TFHE-rs](https://github.com/zama-ai/tfhe-rs), Zama's FHE library. For operation-level latency, use Zama's published numbers: [TFHE-rs benchmarks](https://docs.zama.ai/tfhe-rs/tfhe-rs/1.5/get-started/benchmarks).

Read them as the cost of the raw operation on reference hardware. End-to-end latency in CoFHE adds the steps around the operation: event pickup, validation, and commitment posting.