Accelerating the future.
Today.

Realm is a research laboratory and engineering company. We work from first principles, publish the evidence, and turn difficult systems research into tools people can build with.

Oa Library

One library. Across the stack.

OA (One API, Open Architecture) is a unified GPU-first Rust library with native Python bindings for compute, Ml, Vision, Audio, Media, Data, Crypto, Render, Ui, and Plot through one explicit Vulkan runtime.

Research Laboratory

We study problems that shouldn't be solvable. Then we solve them.

Realm is a research laboratory and engineering company. We work from first principles, publish the evidence, and turn difficult systems research into tools people can build with.

For Developers · OA Rust preview

Rust at the core. Python without a second runtime.

Rust is now OA's primary implementation and API authority. Python mirrors its lowercase modules and PascalCase values: oa::matrix::mat_mul_nt becomes oa.matrix.mat_mul_nt.

The Python preview exposes Engine, Matrix, the functional Ml spine, and Image and Audio surfaces through the same native Vulkan runtime. Remaining module bindings follow the Rust contracts as those surfaces are admitted, without reviving the old Fn* hierarchy.^[*]

Metric	Rust	C++	Python	Result
Wall time / step	7.41 ms	9.19 ms	9.796 ms	Rust −19.4% vs C++
Training throughput	8.63K sample/s	6.96K sample/s	6.53K sample/s	Rust +24.0% vs C++
Token throughput	138.10K token/s	111.38K token/s	104.53K token/s	Rust +24.0% vs C++
Final accuracy	92.4%	92.7%	92.6%	All pass
Generation + reload	Pass	Pass	Pass	Quality parity

1use oa::{Engine, matrix};
2
3let engine = Engine::new()?;
4let one = matrix::ones(&engine, [2, 3])?;
5let two = matrix::full(&engine, [2, 3], 2.0)?;
6let total = matrix::add(&one, &two)?;
7
8assert_eq!(total.read_f32()?, [3.0; 6]);

One spelling system.

Lowercase modules. PascalCase types. snake_case operations. Callers can use ordinary aliases such as import oa.core as oac without OA publishing a second API.

Runtime · Engine

One engine. Explicit execution. Vulkan native.

One owner for device, memory, queues, kernels, scheduling, and profiling. Every module borrows the same runtime and completes through explicit events. OA 0.11.0 is a public development preview. Hardware qualification and performance evidence remain scoped to each named device, driver, and workload.

Realm Inc.

Built to be verified.

We work from first principles and publish the evidence behind every claim we ship. Difficult systems research becomes tools people can build with.

A winged figure reaching toward a beam of light

[*]
One first Release run per language of the same 300 step Char Transformer: batch 64, sequence 16, all position training, evaluation, generation, and fresh checkpoint reload. Measured on Intel Iris Xe TGL GT2 with Mesa 26.2.2 and Vulkan 1.4.354. This is a live first run snapshot, not a fixed clock, multi process performance claim.