| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| 3.1.0 source code.tar.gz | 2026-05-11 | 2.6 MB | |
| 3.1.0 source code.zip | 2026-05-11 | 2.6 MB | |
| README.md | 2026-05-11 | 1.2 kB | |
| Totals: 3 Items | 5.2 MB | 0 | |
Wllama version 3.1 is out - with WebGPU support 🚀🚀
Continuing from the big V3.0 release, V3.1 continues to bring more interesting features into wllama. This release marks 2 major changes:
- 🚀 WebGPU support --> further speed up multimodal processing
- 🤏 Single WASM pre-built binary (no more single/multi-threaded build)
Looking for a demo? --> https://huggingface.co/spaces/ngxson/wllama (Work best on Chromium-based browsers)
View full release note here:
- V3.1: https://github.com/ngxson/wllama/blob/master/guides/intro-v3.1.md
- V3.0: https://github.com/ngxson/wllama/blob/master/guides/intro-v3.0.md
What's Changed
- Use single build for both single-thread and multi-thread by @ngxson in https://github.com/ngxson/wllama/pull/214
- Initial webgpu support by @ngxson in https://github.com/ngxson/wllama/pull/215
- change CONFIG_PATH 'wllama.wasm' to 'default' by @ngxson in https://github.com/ngxson/wllama/pull/216
- V3.1 release by @ngxson in https://github.com/ngxson/wllama/pull/217
Full Changelog: https://github.com/ngxson/wllama/compare/3.0.0...3.1.0