Jede wichtige Veröffentlichung, nebeneinander: neue Versionen der besten Open-Source-KI-Tools — Modelle, Runner, Agenten, Bildgeneratoren — und jedes neue Modell, das in den kommerziellen APIs erscheint. Täglich aktualisiert.
Lade…
Jan v0.8.8: security fixes, llama.cpp 0.7.0, and Trusted Hosts for the local API · A security update to v0.8.7. We recommend updating. Your data and settings ca
Overview · This release adds the LiquidAI LFM3 family (text, vision, audio and omni) together with a new experimental streaming-audio API in mtmd, and wires rea
2.45.5 (2026-10-28) · Bug Fixes · * core: Keep the ExpressionError name in minified builds (#42309) (b3192c6) * core: Keep undefined values in QuickJS expressio
Breaking Changes none · New Features none · Fixes * fix(backups): add missing basic auth to remaining handlers by @tsmith023 in https://github.com/weaviate/weav
What's Changed ⚠️ Compatibility Notes * Default CodeMode to dynamic_catalog=True, so a revealed tool is appended instead of rewriting the cached prompt by @Douw
Core · Bugfixes - Failed TUI worker requests now return an error instead of leaving the caller hanging. (@Nowaker) · Thank you to 1 community contributor: - @No
Feature · * pdf: Run the heading style signal without generate_parsed_pages and add --heading-hierarchy (#5071) (6a12963) * ocr: Replace broken PDF text layers
Hexagon HTP · Microsoft-signed HTP catalog (llama.cpp @ 19cce6). No cert import required on Windows on Snapdragon. · What's Changed * fix(sdk): support flattene
Patch Changes · - Updated dependencies [86ab99c] - @arizeai/[email protected]
1.851.0 (2026-10-28) · Features · * app conversations through the memory_id query parameter (#12025) (07d4f7f) * build, test and run standalone ai agents from a
What's Changed This release fixes 17 security advisories and several smaller issues. In addition, it provides some smaller updates and improvements: * fix build
Signed and notarized PuppyOne Desktop release. · Commit: 9c500b3d88a3bda35b77acf05d9cf3637631eec4 Promoted Internal release: v0.4.0-internal.51 R2 version prefi
Liquid AI LFM3 · LFM3 is Liquid AI's new family of small models built for local agents, tool calling, and structured output, on Liquid's new tri-hybrid architec
What's Changed * model_prefetch: syncrhonize all graph cleanups by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/17178 * Add rope scaling support to M
What's Changed · Bug Fixes - Fix failing flaky tests · Documentation - Update snapshot and changelog for v1.15.31 · Contributors · @lorenzejay
Axolotl v0.21.1 Release Notes · 45 commits since v0.21.0 (October 16, 2026). · This release makes long LoRA runs on large MoEs like Beam smoother, and includes
🔥 What's New? · ⚠️ Breaking Changes · - #15970 - Self-hosted Docker Compose generates its own service passwords per install and won't start without them: run ma
1.34.0 (2026-10-27) · What's Changed Features * feat(verify): stand in for DigitalOcean so Managed Agents can be verified by @enyst in https://github.com/OpenHa
Changelog · New Features: · - Harness Agents in Teams and Workflows: ClaudeAgent and CodexAgent can now be members of an Agno Team and steps of a Workflow, incl
Big-AGI 2.1.3 - Doxy Foxy · The docs move into the app, and every reply says how it ended. · The Big-AGI docs and changelog now open inside the app, read live f
The command-line tool for Vespa.ai. · Use it on Vespa instances running locally, remotely or in the cloud. Prefer web service API's to this in production. · See
2026.10.8 · Highlights · - Updates that finish and recover: an ordinary post-update Doctor failure restarts the Gateway instead of leaving it stopped, an orphan
What's Changed Other changes * chore: replace per-PR changelog edits with generated release notes by @cragwolfe in https://github.com/Unstructured-IO/unstructur
What's Changed * feat(version): add sidebar version footer with update indicator by @nathanpride in https://github.com/miurla/morphic/pull/1078 * feat(version):
Verify this release · Same command for litellm-database, litellm-non_root and the docker.io/litellm/* images · What's new in 1.106.5 · 1 PR. 1 changed. No actio
Highlights · - Timeouts for code that can hang: @tool(timeout=...) limits each function tool call, and timeout= on stream subscribers, @observer, BaseObserver,
What's Changed · Features · * feat(ai-gateway): record model, key owner, byte and upstream error facts on generations by @hassiebp in https://github.com/langfus
Known issues · - Going back to v2.2.4 needs no restore. This release does not change the database. - Instagram shows what a logged-out visitor sees. A profile o
Built-in LLM for Docker · AnythingLLM Docker now has a default local LLM experience right out of the box. The same kind of built-in engine that ships with Anyth
Added · - Cloud tasks: Settings > API Configuration has a Local | Cloud toggle when cloud sessions are on. Cloud picks the Cline model new cloud tasks run on, s
Changes since 1.2.17 · * release(checkpoint): 4.3.1 (#9447) * release(langgraph): 1.2.18 (#9445) * fix(langgraph): let error_handler recover on every path and s
Release note is coming soon...
This release adds scheduled tasks, FLUX-3 edit support, perf improvements and more. Also new: MiniMax-M3.1-Flash, face touch-ups for images and faster performan
This release keeps the Python agent and its existing models. The Pi migration remains separate. · Changes since the last PyPI release: - Add OpenAI browser tool
New · - New Dagster+ organizations no longer start with a [Dagster Plus] Credit Limit Exceeded alert policy. Only the [Dagster Plus] Run Exceeds 24 Hours policy
> [!WARNING] > Stable-Baselines3 (SB3) v2.9.2 will be the last one supporting Python 3.10 (end of life in October 2026) > We highly recommended you to upgrade t
Added · - 🎯 Load selected skills on demand. Administrators can turn on "Load selected skills on demand" under Admin Settings > General or set "ENABLE_SKILL_LAZY
0.125.1 (2026-10-25) · Bug Fixes · * restore the authentic Discord invite (#11967) (98000c4)
Release: v2.26.6 [skip ci] Patch · - Make: Require kernels that build on GCC 16 and load on macOS 14 (343dc7b) - Fix: Reject a zero count in Python search (f6d6
Paperclip v2026.1023.0 · > Released: 2026-10-23 · Paperclip v2026.1023.0 carries 200 commits, promoted from 2026.1012.1-beta.0. The headline is private work: a
Patch release v4.12.1. Security fix only. · Security · - Hardened the external URL policy for user-supplied URLs (chat vision image/audio/video URLs, image and
LoopX v2.1.1 · > Published: The Python package on PyPI, GitHub packages, desktop build assets and signed macOS updater have been independently downloaded and ve
Dify 1.18.0 brings Agents that route workflows, portable app packages, deeper visibility into nested workflow runs, and the opt-in Agent Knowledge Base experien
What's Changed * fix(v3): set additionalProperties false on nullable act action schema by @miguelg719 in https://github.com/browserbase/stagehand/pull/3150 * fi
✨ Features · - Automatic thinking level operation for improved model reasoning #12237 - Resume failed or cancelled delegate subagents #13015 - Subagent session
Updated dependencies to address recent software ecosystem vulnerabilities. (#66023, #65963)
2.13.0 (2026-10-22) · Highlights · This release stabilizes core features while introducing TypeSafe workflow classification, live audio interactions, MongoDB in
Hermes Agent v0.22.5 · Release Date: October 22, 2026 · > Patch release. This tag rolls up the ~81 PRs merged since v0.22.4 into a stable release for the GitHub
Hotfix on top of 0.92.3 · * chore(docker): upgrade Node.js to 24.22.1 · Fixes the Node.js October 2026 security vulnerabilities.
GPUStack v2.3.0 · 🚀 Model Catalog Updates · LLM: DeepSeek-V4.1-Flash-1017, MiMo-V2.6-Pro-MOPD, GLM-5.3-Flash, Qwen3.8-27B, GLM-5.3, Muse-Glimmer-30B, Kimi-K3. ·
Security: eMudhra intermediate CA compromise · - \certifi>=2026.10.19\ is now required. It drops trust in the compromised eMudhra/emSign CAs. - Generated Docker
Changes since langchain-core==1.6.13 · release(core): 1.7.0 (#41418) feat(core): emit tool-call-delta for streamed tool-call args (#41257) · Notes · - Tool-call
What's Changed ✨ New Features * feat(admin): add team and user administration API and CLI by @erichare in https://github.com/langflow-ai/langflow/pull/14782 * f
Changelog · Features 🎶 · - #10750 - BM25 scoring directly on the full-text payload index [docs] - Milestone #57 - TurboQuant: new turbo8 and turbo16 vector data
Patch Changes · - ba7c9a1: Remove the unused @mastra/mcp peer dependency. The provider never imported it, but package managers that install peers automatically
The release significantly speeds up rigid body simulation for scenes with fine convex decomposition or non-convex geometries. Besides, the overhead of trajector
Version 3.17.0 (October 22, 2026) · Significant new features · 1. c00ed8153 Added a batch API in mujoco/mjbatch.h (in Python as mujoco.batch): many simulations
* Bug fixes * Built C++-generated intrinsic LLVM IR on Windows with clang-cl instead of emitting empty bitcode, enabling vectorized CPU intrinsics on Windows (h
🔒 Security Notes · - Haystack now requires certifi>=2026.10.19. certifi is installed transitively through httpx and posthog; earlier versions trust eMudhra/emSi
> Full technical notes for this release: docs.lykos.ai/stability-matrix/release-notes/2.16.5 Added - Added Fizgig, a LoRA training studio for Flux 2 Klein 9B, K
What's Changed * increase chunk size from 512 bytes to 64 KB for HTTP response reading by @YasirAlibrahem in https://github.com/microsoft/markitdown/pull/1511 *
This release adds model merging for all four model types, reduces redundant Hub requests when loading models, and distributes evaluator inference across DDP ran
MLflow 3.17.2 is a patch release that contains a minor feature enhancement and bug fixes. · Features: · - [Evaluation] Record dataset, scorer, and agent lineage
Added · - \[Server API\] The initial_replicas parameter of the task creation request (POST /api/tasks) sets the number of replica jobs created for each annotati
Changelog · 🚀 Features execution - daa30bf add execution and taskrun state inside OTEL spans · traces - 4e9be71 add taskId inside OTEL spans · 📘 Subtasks core -
Isaac Lab 3.0 · Isaac Lab 3.0 is now generally available. · v3.0.0 is the official Isaac Lab 3.0 release and the recommended baseline for new projects and migra
Highlights · *950 PRs from 245 contributors.* · New models · | Model | Type | Cookbook | |---|---|---| | Clef / Clef Flash | LLM / VLM | link | | PPLX-Decider v
📦 Release v2.2.20 · This release was automatically published from PR #21078. · Changes See PR description: https://github.com/lobehub/lobehub/pull/21078 · Commi
v0.32.0 · Highlights · This release features 814 commits from 363 contributors (121 new)! · * Structured decisions and generic parsers: the new POST /v1/systemo
Lyrical sync Oct 16
1.0.0-rc2 is the second release candidate for v1.0.0. This release fixes several issues in rc1 and introduces new features and expanded support. · Highlights 1.
Features · 🛠️ TRL now trains on AWS Trainium · · The fused LM head from v1.15 is what made Neuron tractable; the remaining gaps are closed here. There is a guid
Nanobrowser v0.3.0 · Nanobrowser 0.3 is rebuilt from the ground up: one agent in the side panel, on top of a new browser control layer that clicks the right ele
KTransformers v0.7.2: DeepSeek-V4-Flash LoRA Fine-Tuning · KTransformers v0.7.2 adds LoRA fine-tuning for the April DeepSeek-V4-Flash checkpoint. Routed experts
What's Changed · Version bumped from 1.0.55 to 1.0.56 · Recent Changes - Skip DataDragon ledger DDL when the schema is current so concurrent runs stop deadlocki
🎉 Crawl4AI v0.9.5 Released! · 📦 Installation · PyPI: · Docker: · Note: Docker images are being built and will be available shortly. Check the Docker Release wor
Plane-4.7.2 was released as the stable/official version for planes and VTOLs on 11-Oct-2026 Official release announcement: https://discuss.ardupilot.org/t/plane
koboldcpp-1.123.1 · · - NEW: Introducing koboldcpp.net, our new community website! - KoboldCpp now has an official community website! - This is a legitimate com
Agent Zero v2.14 · Highlights · - Significantly faster startup and chat loading — Embedding warm-up now overlaps with chat restoration and startup watchdogs, re
v4.6.0 · JEV Support (typed-decision / System-1) · - Added the decision task type for JEV-style decision models: instead of generating text, the model returns c