Runtimes & apps Published:
Open WebUI 0.12.4 loads skills on demand, compacts more chats and ships security fixes
What happened
Open WebUI v0.12.4 adds an optional "Load selected skills on demand" setting (ENABLE_SKILL_LAZY_LOADING, off by default) that only lists selected skills for the model to load when needed, instead of sending their full instructions with every message. Context compaction now also runs in temporary chats and in short chats already over the threshold, so follow-ups no longer fail with a context length error, and a "Context compacted" marker shows where it happened. Administrators can also hide individual controls from users, and MCP connections now support the stateless 2026-07-28 protocol. The release contains security fixes the project has not listed yet and strongly recommends updating as soon as possible.
Why it matters locally: Anyone running a self-hosted Open WebUI should update for the security fixes, and teams on local models with smaller context windows get shorter prompts and fewer context length errors.
Three ways to read it
For the market
Admin features like per-control hiding, LDAP email defaults and token usage in audit logs suggest Open WebUI is aiming at company-wide rollouts, where it competes with hosted chat workspaces. Support for the newer MCP protocol likely helps it keep up with newer tool servers.
In plain words
Open WebUI is a free chat app for AI that many offices run on their own server. This update patches security holes, and it stops long conversations from suddenly failing by quietly summarizing older messages, like a secretary who keeps short minutes instead of the full transcript.
For practitioners
Update first for the unlisted security fixes. If you attach skills to models, try ENABLE_SKILL_LAZY_LOADING (it needs the built-in Skills tool), since 0.12.3 started sending full skill instructions with every message; check that your local model actually calls the Skills tool when it needs one. Tool definitions from tool servers now reach the model in the same form after a restart, so prompt caching keeps working, and UI_VISIBILITY lets you hide controls without changing their values.
Sprout's insight
Sprout compares this with other brands and with where real applications are heading. Free for members.
Sign in free to read it →Subscribe to Sprout's weekly
One email a week: the top stories and a taste of Sprout’s insights. Unsubscribe any time.
Source
Read the official source: Open WebUI v0.12.4 release notes (GitHub) ↗Summaries are written by OnPrem World from the official source. We link to the original and never republish it.
More news
- MiniMax publishes M3.1-Flash weights: a 428B multimodal MoE with a 1M-token context and a 252 GB download
- NVIDIA's small Nemotron 3.5 Bolt E4B arrives in Ollama 0.40.10, with a DFlash draft for faster replies
- Ollama 0.40.9 adds local text-to-speech on Apple Silicon, d3 decision models, and Flint on Linux and Windows
- Recursive releases Flint, an Apache 2.0 coding model with 40B parameters and about 2.5B active, already in Ollama for Mac
- AMD and Intel agree on oneDNN as PyTorch's shared x86 CPU acceleration library