lemonade-sdk/lemonade candidate-v2026.44.0

v2026.44.0 (release candidate)

on GitHub

Headline

⚠️ These notes are AI generated and will be revised by a human ⚠️

• lemonade launch pi now supports MCP mode (--mcp) and code mode (--codemode), with per-model context windows, image input, and tool-calling.
• sd-cpp ROCm backend is re-enabled on Windows with a full HIP runtime bundle, enabling model streaming on AMD integrated GPUs.
• Auto-tuned context sizing now reserves runtime buffer memory for discrete GPUs, unified-memory GPUs, and host backends, improving model fit accuracy.
• LFM2.5-24B-A2B-GGUF joins the model registry with labels for chat, reasoning, and tool-calling.

Breaking Changes

⚠️ These notes are AI generated and will be revised by a human ⚠️

• The /v1/models/{model}/options endpoint's resolved_ctx_size field can now return null when automatic GPU sizing cannot fit a valid context; clients must handle null gracefully.

Lemonade Server

Other platforms? See our Installation Options for Docker, Snap, Arch, Debian, and more.

Embeddable Lemonade

Portable binaries for bundling into your own installer. Run lemond ./ as a subprocess.

What's Changed

Thanks @QuTosh, @bernardogv, @bitgamma, @fl0rianr, @iswaryaalex, @jeremyfowers, @joshuacox, @meghsat, @sawansri, @sreeram-11, @superm1, @zaneni6 for your awesome contributions to this release!

New Contributors

• @joshuacox made their first contribution in #3784
• @bernardogv made their first contribution in #3895

Full Changelog: v2026.42.1...candidate-v2026.44.0

The Windows installers in this release candidate are not signed; Windows SmartScreen will warn before installing. Stable releases are signed.