AI.info
AI Tools Updates — Page 119
What changed in the AI tools this week — dated, from the makers’ own changelogs.
- DeepEval — Introducing Jev in DeepEval
DeepEval 4.2.2 introduces support for Jev, TypeSafe AI's System One model.
- DeepEval — typescript-v0.9.17
New release.
- Harness AI (AIDA) — Dynamic AI Discovery & Runtime Security for AgentCore
A new integration adds continuous visibility and real-time threat detection across agent interactions on AWS.
- Together AI — Canary rollouts: upgrade models in production without downtime
Dedicated inference supports staged traffic ramps, metric gates, and automatic rollback for model upgrades.
- Letta — Release v0.32.16
Set the cloud backend as the default at startup.
- Typeface Arc — The Next Generation of the Typeface Orchestration Engine: Closing Marketing's Open Loop
Closing Marketing's Open Loop with the Typeface Orchestration Engine.
- Mistral Vibe — A simpler Vibe experience
Chat and Work are merging into one Vibe experience, with fewer decisions and less navigation.
- Mistral Vibe — Mini App Canvas
Create an app you can review, edit, and share without writing code or leaving the conversation.
- Mistral Vibe — Agentic Knowledge Base
Vibe Work now builds and maintains a structured personal knowledge base for you.
- Mistral Vibe — Spreadsheets in Vibe Work
Work with Excel and CSV files in Vibe Work, from analyzing uploaded data to creating a workbook from a prompt.
- Instantly — The new Instantly Agents are here
Agents find people, build messages, and start conversations.
- PixAI — PixAI Weekly Update (Sep 15 – Sep 22)
Tsubaki.3 is fully live; Photo Fail Generator and LoRA advanced training are live; the Editor was upgraded to Repaint.
- Higgsfield AI — GPT Image 2.5. Now on Higgsfield
GPT Image 2.5 is available in Flare and Sunburst variants.
- Higgsfield AI — GPT-6 Sol and GPT-6 Luna. Now in Supercomputer and MCP
GPT-6 Sol and GPT-6 Luna are now available in Supercomputer and Higgsfield MCP.
- Higgsfield AI — Seedance 2.5 Draft Mode. Preview before the final
Seedance 2.5 Draft Mode adds 480p previews before generating a selected take at 1080p with Seedance 2.5.
- Tabnine — v6.6.2
Fixed requests made with an expired or invalid sign-in token returning a server error instead of prompting you to sign in again.
- Tabnine — v6.6.1*
Administrators working with very large user lists no longer hit request size limits.
- Darktrace DETECT — Darktrace Extends Behavioral Security to Help Customers Safely Adopt Enterprise AI with General Availability of Darktrace / SECURE AI™
- True Fit Fit Intelligence Platform — True Fit + Glance Bring Fit Intelligence Into the Agentic Shopping Experience
Bring Fit Intelligence Into the Agentic Shopping Experience
- Deepgram — Nova-3 Improved Models for Danish, Estonian, Flemish, Italian, Lithuanian, Macedonian, Polish, Urdu, and Vietnamese
Released improved Nova-3 monolingual models for nine existing languages, enhancing transcription quality for batch and streaming workloads.
- Wordly — Wordly for Microsoft Teams 2.3.1
We've dynamically increased the height of the audio window and fixed some issues with sign in.
- v0 by Vercel — More model choices, with Enterprise owner controls
Browse and search AI Gateway models through Add models in the chat composer.
- Factory Droids — CLI v0.225.0, Desktop v0.182.0: Default app file opening, subagent controls, Droid Shield accuracy, and desktop sign-in and menu fixes
Added default-app file opening and controls to stop subagents; changed Droid Shield checks and Git marketplace installs.
- Honeycomb — More Canvas Connectors
Canvas Agents now connect to Amplitude, Atlassian, Braintrust, Checkly, incident.io, Notion, Pylon, and Sentry.
- Google AI Studio — Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS generally available (GA)
Released two text-to-speech models and the Gemini API Voices endpoint; introduced voice design, voice replication, and Extended Voice Library.
- Omni — September 14, 2026
Admins can disable delivery destination types, and MCP tools now support reading and editing app drafts and published apps.
- Weaviate — v1.38.17 - LSM store performance improvements
- Weaviate — v1.39.6 - LSM store performance improvements, HNSW ACORN & HFresh index Fixes
Adds a Weaviate Embeddings rate limit and removes the /v1/modules endpoint and contextionary extension/concept handlers.
- Snyk Open Source — Evo MCP Server now available
Evo's MCP Server lets MCP clients query AI assets, connections, policies, and policy violations, and create or update policies.
- Gemini Enterprise Agent Platform — Anthropic's Claude Opus 5.5
Claude Opus 5.5 is available in Model Garden.
- Tines Workflow Automation — 🌐 Create and delete tunnel networks via API
Create and delete tunnel networks via API.
- Tines Workflow Automation — 🔥 New pfSense connector
The pfSense connector is available.
- Kimi — 3.2.12 (2026-09-22)
Select content in PPT, Excel, Word, or Markdown documents and ask Kimi to make precise edits; attach files of any size.
- Ask Sage — v3.9.1
Claude Code 502 fix; Sign in with Ask Sage fix
- Distro — 📅 Past and upcoming meetings in one view
The Meetings report now shows past and upcoming meetings together in one date range.
- Devin — September 21, 2026
SWE-2 is rolling out as a research preview in the agent selector, and you can attach a folder to a message.
- Devin — September 4, 2026
Lightboxes support arrow-key navigation and Mermaid diagrams; PR comments are combined, and Code Scan adds multi-select and no-op runs.
- UiPath Maestro — Cartographer now feeds into the Model – Design executable processes stage
Cartographer now sits upstream of the Model – Design executable processes stage and generates a Process Design Document and Specification Design Document.
- Concierge — 📅 Past and upcoming meetings in one view
The Meetings report shows past and upcoming meetings together in one date range.
- incident.io On-call — Investigations are 2x faster
The median time from an incident opening to an accurate channel message went from 6.7 to 3 minutes.