AI.info
AI Tools Updates — Page 20
What changed in the AI tools this week — dated, from the makers’ own changelogs.
In inglese
- Sierra Insights — Next gen multimodal agents: an interface that morphs with the conversation
Sierra’s multimodal agents bring voice, text, and visuals into the same conversation.
- RunPod Pods — Sales tax and tax ID support
Runpod collects sales tax on credit purchases. Add a business tax ID at checkout or in your account settings.
- Deepgram — @deepgram/react 0.2.0
Adds runtime Voice Agent controls and makes React session lifecycle handling safer.
- v0 by Vercel — Team chats are easier to discover and collaborate on
New chats created in a team workspace are now visible to the team by default.
- Microsoft Copilot — Build apps in Copilot Cowork and Copilot Studio
Copilot Cowork and Copilot Studio let users create full-stack business apps through conversation, grounded in their data.
- Digits AI Invoicing — New Accountant Role for Firms
Firms can assign an Accountant role with client-level permissions except bill payments and invoice collection.
- Mixedbread Rerank — MCP Tools in the Responses API
You can define MCP servers in the Responses API so Toast 1 can search using external MCP tools.
- Auphonic — Editable Gain Curve: Refine the Adaptive Leveler in the Auphonic Editor
Refine the Adaptive Leveler in the Auphonic Editor with an editable gain curve.
- Dataiku (DSS) — Version 15.0.1 - September 10th, 2026
Added per-user custom instructions; Cobuild can create and manage ML evaluation recipes and Agent evaluation recipes, and inspect, create and edit Dashboards
- Agent Studio — Next gen multimodal agents: an interface that morphs with the conversation
Sierra’s multimodal agents bring voice, text, and visuals into the same conversation.
- Truss — Regional deployments
Restrict every replica in a deployment to a specific Baseten region. Select a region through the dashboard, Baseten CLI, or Management API.
- Truss — DeepSeek V4.1 Flash available on Baseten
DeepSeek V4.1 Flash is now available through Baseten Model APIs.
- AI Dungeon — Voyage Creator API: Now Open to All
Voyage Creator API is now open to all.
- Intercom Fin — Track data connector health over time
Status codes now appear as a chart over time alongside the existing latency chart.
- Intercom Fin — Export a saved View exactly as you see it
Export any saved Inbox View to CSV as a point-in-time snapshot, with the same columns, sort order, and a link to every conversation.
- AssemblyAI — Assembly Agent Now Available in the Dashboard
The AssemblyAI dashboard now includes Assembly Agent, an AI assistant that answers questions and generates code you can run for your use case.
- DeepSeek — DeepSeek-V4.1-Flash Release
DeepSeek V4.1 Flash is available on the API with native multimodal support; legacy model names route to V4.1 Flash.
- Voiceflow — Instruction patching
Agent instructions, global prompts, and playbook instructions can be read, searched, and patched line by line through the API.
- Mixedbread Embed — MCP Tools in the Responses API
The Responses API can use external MCP servers and tools with Toast 1.
- vLLM — v0.29.0
Model Runner V2 is now the default for all models.
- Retool Agents — New AI models available in Retool
Retool added GPT 5.4 Mini, GPT 5.4 Nano, GPT 6 Astra, Claude Fable 5.1, and Gemini 3.8 Flash for AI resource queries.
- Mistral AI — Cloudera and Mistral Partner to Bring Specialized, Sovereign Intelligence to Enterprise Data
Cloudera and Mistral are bringing specialized, sovereign AI to enterprise data.
- Vyond Go — September 2026 Vyond Newsletter
Vyond Turbo can create videos with animated or realistic narrators and automatically generated background music.
- Skyvern — Release v1.0.52
Let Copilot edit pending workflow drafts.
- whisper.cpp — b5127
Windows arm64 CUDA builds use a preview CUDA Toolkit for Windows on Arm and are experimental.
- Visual Copilot — Introducing Agent Native Analytics
Agent-Native Analytics, a tool Builder uses internally, was open-sourced.
- Visual Copilot — Builder Code and Builder Content: same products, same features, clearer names
Fusion was renamed Builder Code, and Publish was renamed Builder Content; functionality stayed the same.
- CodeRabbit — Connections and scopes | Management
Configure external service connections once and control product access with base and named scopes.
- Vectara Platform — Caller Identity Covers Every Principal
session.caller.type now reports one of six values: user, anonymous, api_key, app_client, agent, or service_account.
- Baseten — DeepSeek V4.1 Flash available on Baseten
DeepSeek V4.1 Flash is now available through Baseten Model APIs.
- Baseten — Regional deployments
Restrict every replica in a deployment to a specific Baseten region.
- Riverside.fm — See your hosting analytics on iOS
See your podcast analytics right from the iOS app.
- Hex — More control over Generative app iteration
The app builder now supports direct text editing, creating chart properties, and queuing visualization tweaks.
- Pactum Negotiation Suite — What’s New at Pactum: AI Agents Take On More Procurement Work | September 2026
Pactum's update includes AI agents for direct and indirect procurement, supplier engagement, sourcing, negotiation, and commercial intelligence.
- Retool — New AI models available in Retool
Retool added support for GPT 5.4 Mini, GPT 5.4 Nano, GPT 6 Astra, Claude Fable 5.1, and Gemini 3.8 Flash for AI resource queries.
- Genspark — Introducing Gen-1 Slides: AI Model Built for Work
Introducing Gen-1 Slides.
- Krea AI — Krea Agent is open on every plan, with launch-week free credits
Krea Agent is now available to everyone, including Free; every plan gets the context library, skills and integrations.
- RunPod Serverless — Sales tax and tax ID support
Runpod now collects sales tax on credit purchases in applicable jurisdictions. Add a business tax ID at checkout or in account settings.
- Runway ML — Agent Brand Kits
Agent can now create and edit Brand Kits for you, right in the conversation.
- Claude AI — Smart reports (beta)
We’ve launched smart reports, which analyze how a team uses Claude and report on the work getting done, what it costs, where sessions run into friction, and whi