v3.12.1

mudler/LocalAIv3.12.1Feb 21, 2026by mudler

AI Summary

Patch release to fix incompatibilities with Qwen 3 coder by updating llama.cpp. Includes backend traces feature and drops remaining bark.cpp code.

Key Highlights

Updated llama.cpp to fix Qwen 3 coder incompatibilities
Added backend traces feature for request monitoring
Dropped bark.cpp leftovers from pipelines
Fixed merge logic for openresponses messages
Updated documentation version

New Features

Backend traces for request monitoring
Qwen 3 coder compatibility fix

Full Release Notes

<!-- Release notes generated using configuration in .github/release.yml at master -->

This is a patch release to tag the new llama.cpp version which fixes incompatibilities with Qwen 3 coder.

## What's Changed
### Other Changes
* docs: :arrow_up: update docs version mudler/LocalAI by @localai-bot in https://github.com/mudler/LocalAI/pull/8611
* feat(traces): Add backend traces by @richiejp in https://github.com/mudler/LocalAI/pull/8609
* chore: :arrow_up: Update ggml-org/llama.cpp to `b908baf1825b1a89afef87b09e22c32af2ca6548` by @localai-bot in https://github.com/mudler/LocalAI/pull/8612
* chore: drop bark.cpp leftovers from pipelines by @mudler in https://github.com/mudler/LocalAI/pull/8614
* fix: merge openresponses messages by @mudler in https://github.com/mudler/LocalAI/pull/8615
* chore: :arrow_up: Update ggml-org/llama.cpp to `ba3b9c8844aca35ecb40d31886686326f22d2214` by @localai-bot in https://github.com/mudler/LocalAI/pull/8613


**Full Changelog**: https://github.com/mudler/LocalAI/compare/v3.12.0...v3.12.1