What's new from vendors
Real product updates pulled from each vendor's own changelog page. Fresher than our editorial re-review cycle; grounded in what vendors are actually shipping.
Updated to parity with Claude Code v2.1.298.
Added nextFireAt to Cron tools and property_order to elicitation requests. Fixed rewind_conversation errors and in-process MCP server naming issues. Updated to parity with Claude Code v2.1.299.
Fixed Unicode separator issues in session imports and background subagent reports. Reverted canUseTool options for connector tools. Reduced browser SDK bundle size by 40%. Updated to parity with Claud
Added per-tool permission policies and built-in tool name exports. Fixed setModel() failures at usage limits and disk-full login errors. Updated to parity with Claude Code v2.1.301.
Fixed structured output errors, changed default permission mode for third-party providers, enabled fork subagents by default, and updated to parity with Claude Code v2.1.302.
Send notifications from Loop automations and alerts directly to PagerDuty. Configure integration keys and incident severity for each destination. Requires data plane v2.17.0+.
Patterns assigns severity to findings for prioritization and filtering. Users can correct severity with a reason, which Loop preserves in later investigations. Requires data plane v2.18.0+.
Edit video at the 4K tier with Kling o3 on RunDiffusion. Compare a lounge before and after a navy upholstery edit and read the exact prompt.
Guide Gemini Omni Flash 1.1 video with reference images and clips on RunDiffusion. See a university courtyard reference and its generated camera pass.
Generate video from text or images with Gemini Omni Flash 1.1 on RunDiffusion. See an interior walkthrough and review frame and resolution settings.
Use image and video references with MiniMax H3 Max Reference to Video on RunDiffusion. See a landscape reference and its generated video.
Generate video from text or animate a start image with MiniMax H3 Max on RunDiffusion. Review duration, resolution and start/end frame controls.
Release note for Refinder AI version 1.5.0.
Release note for Refinder AI version 1.7.0.
Release note for Refinder AI version 1.8.0.
Web App v2.8.0 release including workflow organization features, 3D downloads, and platform refinements.
MCP Server v0.4.2 update focusing on Cursor authentication and web interface improvements.
Compute release v0.29.0 adding support for LTX 2.3, Lyria 3 Music, and xAI Video Models.
API update v2.114.0 introducing new models, Tripo AI BYOK support, and general API updates.
Release of MCP Server v0.5.0 featuring rich media visualization and client enhancements.
انتقال ميزة Model Access إلى قسم Manage مع تطبيق قواعد على مستوى المؤسسة، مما يتيح التحكم المركزي في النماذج المتاحة للأعضاء والمساحات.
إضافة خطوة إلزامية لعملاء Enterprise للإبلاغ عن استخدام نماذج المصادر المفتوحة، تتطلب تحديد مساحة العمل التي يتم الإبلاغ منها.
إتاحة واجهة Krea 3 البرمجية في الوصول المبكر مع دعم Zero Data Retention، وإعادة تسمية أوضاع التفكير (fast/high) مع جدول زمني لإيقاف الدعم القديم.
تصبح Krea 3 نموذج الصور الافتراضي، وإضافة التحقق من الرصيد عبر MCP، وإطلاق Topaz Starlight Fast 3، ودعم دقة 2K و4K في FLUX 3 Video.
إطلاق Wan 3.0 Prime للفيديو، وإضافة تقديرات التكلفة قبل التوليد في Agent وMCP، وإصلاحات متعددة لـ Krea 3 وEdit وAgent وiOS.
Addresses security vulnerabilities in OpenTelemetry and WebSocket dependencies, isolates Compose environment and configuration, implements atomic configuration defaults, and updates installation and r
Introduces readable configuration names, performs automatic configuration migration from hashed IDs, makes config_override execution-only, and fixes guided tour image prefetch issues.
Routes configuration overrides through the configuration service, manages engine lifetime with Compose, optimizes engine channel sockets, updates docs and agent skills, and includes migration notes fo
Improves trace query performance, updates Compose dependency handling, adds Linux worker compatibility for glibc builds, handles empty package configuration defaults, and fixes Python SDK connection r
Adds namespace routing in SDKs, rejects out-of-scope namespaces, enforces message size limits, updates Compose progress display, improves installer setup, enhances sandbox isolation, and reduces engin
Adds configurable timezone support (IANA/UTC offset) to fix LLM time calculation errors and standardize timestamps. Introduces 'disableThinking' configuration for various LLM providers to optimize rea
Changelog entry published 4 days ago announcing Python SDK v5 and JS/TS SDK v6.
Changelog entry published 4 days ago about sharing LLM connections across projects.
Changelog entry published 2 days ago regarding permissions for project API keys.
Release candidate for oMLX 0.7.1 adding Muse Spark Open support, model recommendations for Macs, Splash as an external engine, reworked settings/chat, DFlash in batched engines, and decision models.
Exposes cube orderBy in TypeScript SDK types. Fixes bindings to build against wren-core 0.3.2.
Second release candidate for v1.0.0, fixing rc1 issues and introducing Agentic mode, Team/Me sharing, new model providers, and arm64 Docker images.
Patch release adding a 'Custom development' button to the dashboard sidebar that opens a contact form for custom guardrails, hooks, and integrations.
Patch release removing the API Bot announcement banner and the API Bot sidebar card from the dashboard.
Reflection’s first model, Beam, scores 30 on the Artificial Analysis Intelligence Index, placing it #2 among US open weights models.
Support emails now arrive in the SiteSpeakAI Inbox alongside chats, with AI-generated answer drafts to help your team respond faster.
Cloud agents can now build interactive canvases. The Environments API is in beta for programmatic management. Environments support up to 300 repositories and load 5x faster.
Adds StarRocks as a first-class data source. Fixes include counting errors in validate warnings, keeping cubes during MDL import, and resolving profile secrets from the selected project.
Hand off a brief to OpenAI, follow the work in Hamster, and get the code back for review. Hamster now reads and ships to your GitLab projects, too.
Nvidia’s Nemotron 3.5 Super VL is now the leading US open weight model, scoring 26 on the Artificial Analysis Intelligence Index.
Supports pluggable access control providers and adds StarRocks as a first-class data source. Fixes include keeping explicit ROWS window frames and preserving string literals.
Launch of AA-Robotics v0.1, a research preview measuring frontier language model performance on understanding, planning, and acting in the physical world using a 6-axis robot arm.
Patch release fixing Claude thinking text surfacing, Anthropic refusal stop reporting, Bedrock GPT function tool handling, and Playground request headers.
Adds StarRocks as a first-class data source.
Upgrades DataFusion to 55.1.