Baixar Lemonade Server 2026.43.3 per a Windows

Headline
- TheNoise image backend promoted to Core tier for ROCm on AMD GPUs across Linux and Windows.
- Image editing now accepts multiple reference images with automatic size selection and lossless passthrough.
- Live inference metrics added to GET /v1/stats, showing active requests and estimated tokens per second.
- AMD GPU utilization and VRAM telemetry now reported on Windows alongside NVIDIA in system-stats.
Breaking Changes
- GET /api/v1/system-stats now reports gpu_percent and vram_gb for the primary GPU by total memory instead of the GPU with the highest utilization — on multi-GPU systems these fields will show different values, requiring updates to dashboards and alerting rules.
- Models gated behind a guest-tier backend are now hidden from GET /v1/models?show_all=true and return model_not_supported on pull until the guest backend is installed — existing setups relying on those models being visible will no longer see them.
What's Changed
Thanks @bitgamma, @fl0rianr, @jeremyfowers, @joshuacox, @sreeram-11, @zaneni6 for your awesome contributions to this release!
Click to expand changelog
- refactor(server): run backend servers through ServerProcess and NativeProcess by @jeremyfowers in https://github.com/lemonade-sdk/lemonade/pull/3705
- fix(server): support multi-reference image editing by @fl0rianr in https://github.com/lemonade-sdk/lemonade/pull/3801
- fix(config): read the distro defaults override from lemond's own install prefix by @jeremyfowers in https://github.com/lemonade-sdk/lemonade/pull/3775
- feat(images): add per-model edit size caps and Qwen-Image-2.1 Turbo INT8 by @fl0rianr in https://github.com/lemonade-sdk/lemonade/pull/3803
- fix(models): preserve catalog identities in extra_models_dir by @fl0rianr in https://github.com/lemonade-sdk/lemonade/pull/3805
- fix(models): prevent duplicate discovery of incomplete shards by @fl0rianr in https://github.com/lemonade-sdk/lemonade/pull/3811
- docs: add systemd start step to the Arch install page by @joshuacox in https://github.com/lemonade-sdk/lemonade/pull/3784
- ci: bump auto-labeler to Claude Haiku 5.5 by @fl0rianr in https://github.com/lemonade-sdk/lemonade/pull/3774
- feat(backends): label each backend with a tier and a format by @jeremyfowers in https://github.com/lemonade-sdk/lemonade/pull/3706
- Update Lemonade to ROCm 10.1.0 by @sreeram-11 in https://github.com/lemonade-sdk/lemonade/pull/3821
- chore(flm): bump FLM to v1.0.8 by @zaneni6 in https://github.com/lemonade-sdk/lemonade/pull/3810
- Build cpp-httplib once instead of in every file that includes it by @jeremyfowers in https://github.com/lemonade-sdk/lemonade/pull/3837
- fix(server): report NVIDIA GPU and VRAM usage by @fl0rianr in https://github.com/lemonade-sdk/lemonade/pull/3282
- mark TheNoise as Core-tier by @bitgamma in https://github.com/lemonade-sdk/lemonade/pull/3826
- test: pre-pull server_llm.py's models and cover implicit pull on inference by @jeremyfowers in https://github.com/lemonade-sdk/lemonade/pull/3839
- lemonade bench: add prefill TPS by @bitgamma in https://github.com/lemonade-sdk/lemonade/pull/3759
- Add AnythingLLM and Phoronix Lemonade 11.9 articles to news page by @jeremyfowers in https://github.com/lemonade-sdk/lemonade/pull/3836
- fix(server): expose live inference throughput across backends by @fl0rianr in https://github.com/lemonade-sdk/lemonade/pull/3791
- test: isolate every C++ test from the real config dir by @jeremyfowers in https://github.com/lemonade-sdk/lemonade/pull/3612
New Contributors
- @joshuacox made their first contribution in https://github.com/lemonade-sdk/lemonade/pull/3784
Full Changelog: https://github.com/lemonade-sdk/lemonade/compare/v2026.42.1...v2026.43.3

Windows installers are signed. Free code signing provided by SignPath.io, certificate by SignPath Foundation. See our Code Signing Policy.