v0.5.2

Full Changelog: v0.5.1...v0.5.2

v0.5.1

What's Changed

• Code improvements (#21–#25) by @m96-chan in #26

Full Changelog: v0.5.0...v0.5.1

v0.5.0

What's Changed

• Support additional BitNet models (Falcon-E, BitNet 0.7B) by @m96-chan in #14
• Optimize inference: fused kernels, GEMV loop, fix uniform leak by @m96-chan in #15

Full Changelog: v0.4.0...v0.5.0

v0.4.0

What's Changed

• Add oxbitnet-ffi crate — C FFI bindings by @m96-chan in #9
• Add oxbitnet-java crate — Java/JNI bindings by @m96-chan in #11
• Add oxbitnet-swift — Swift bindings via C FFI by @m96-chan in #12
• Add publish workflows for Swift (SPM) and Java (Maven Central) by @m96-chan in #13

New Contributors

• @m96-chan made their first contribution in #9

Full Changelog: v0.3.1...v0.4.0

v0.3.1 — Rust & Python Bindings

What's New

Rust & Python bindings (#3)

0xBitNet is now available in three languages:

Package	Install
0xbitnet (npm)	npm install 0xbitnet
oxbitnet (crates.io)	cargo add oxbitnet
oxbitnet (PyPI)	pip install oxbitnet

Rust — Full BitNet inference via wgpu (Vulkan/Metal/DX12). Streaming token generation with impl Stream<Item = String>.

let mut model = BitNet::load("model.gguf", Default::default()).await?;
let mut stream = model.generate_chat(&messages, options);
while let Some(token) = stream.next().await {
    print!("{token}");
}

Python — PyO3 bindings with streaming callback support.

model = BitNet.load_sync("model.gguf")
model.chat(
    [("system", "You are a helpful assistant."), ("user", "Hello!")],
    on_token=lambda t: print(t, end="", flush=True),
)

Other changes

• Interactive chat examples for Rust (cargo run --example chat --release) and Python
• Unified CI: v* tags trigger npm + crates.io + PyPI publish in parallel
• Rust CI with clippy and tests on every push to packages/rust/

Fixes (v0.3.1)

• Fix macOS x86_64 wheel build (macos-13 runner deprecated)
• Fix npm provenance (missing repository in package.json)

v0.2.0 — Node.js WebGPU support

What's New

Node.js CLI Example (#2)

Run BitNet from the command line using Node.js and the webgpu npm package (Dawn bindings).

cd examples/node-cli
npm install && npm start

• Interactive chat with streaming token output
• Progress bar for model download/parse/upload
• --url, --max-tokens, --temperature flags
• tok/s performance metrics

Documentation

• README: added Node.js CLI example, updated platform support and architecture tree
• Getting Started guide: new Node.js Usage section with full setup code
• API Reference: clarified LoadOptions.device for Node.js usage

Platform Support

• Node.js: confirmed working via webgpu npm package (Dawn)
• Deno: confirmed working with --unstable-webgpu

Full Changelog: v0.1.3...v0.2.0

v0.1.3

Fix hero image URL in npm README (use absolute GitHub raw URL).

Full Changelog: v0.1.2...v0.1.3

v0.1.2

Include README in npm package.

Full Changelog: v0.1.1...v0.1.2

v0.1.1

Changes

17bb307 Release v0.1.1
b1fa218 Narrow supported models to BitNet 2B-4T only, use official Microsoft GGUF URL
85a29f6 Add back navigation to examples index and tighten TL;DR summary prompt
c327472 Add CNAME for 0xbitnet.m96-chan.dev custom domain
19cd991 Fix preset model URL to official Microsoft GGUF
291763d Add preset model selector with BitNet 2B-4T to both example apps
5665a12 Mark BITNET_3B_CONFIG as deprecated (upstream model removed)

Full Changelog: https://github.com/m96-chan/0xBitNet/commits/v0.1.1