AI.info
AI Tools Updates — Page 19
What changed in the AI tools this week — dated, from the makers’ own changelogs.
In inglese
- Recraft — New Model: ChatGPT Images 2.5
Available in Recraft Studio for image generation and editing, with 50% lower latency than its predecessor.
- vLLM — proto-v0.1.0
Released vllm-proto 0.1.0.
- Profound — Microsoft 365 Integration, Personal Context, and more in Profound
Agent Analytics now accepts log delivery via S3 alongside the Custom Log API.
- Sigma — Connect multiple Sigma organizations to one Slack workspace
You can now connect multiple Sigma organizations to the same Slack workspace.
- whisper.cpp — v1.9.4
- Mintlify — Simplified credit pricing, OpenAPI Overlays, and more
Assistant answers use 25 credits, successful automation runs use 250, and editor and Slack agents are free on Pro and Enterprise; OpenAPI Overlays can transform specs.
- Vectara Platform — Duplicate Document IDs Are Rejected Before Anything Is Indexed
core_document_index now rejects a batch in which two entries resolve to the same document ID, before any document is deleted or indexed.
- Vectara Platform — Ingest Multiple S3 Prefixes in One Pipeline
An S3 pipeline source scopes ingestion with prefixes, a list of up to 100 key prefixes.
- Vectara Platform — Pipeline Sources Can Propagate Record Deletions
A pipeline source can report that a record was removed, not just created or updated.
- Vectara Platform — Spreadsheet, PowerPoint, and Word Libraries in the Sandbox
The sandbox runtime now ships document and spreadsheet packages alongside numpy and pandas.
- CoreWeave Cloud — New Billing
The FOCUS Billing Export API is now available in public preview. Pull hourly CoreWeave usage in FOCUS 1.2 JSON or CSV format.
- Neon — Claimable Neon, @neon/tools, and more release - Sep 11, 2026
Claimable Neon is provisioned by agents and claimed by humans.
- Baseten — Model API costs
GET /v1/billing/model_apis returns exact subtotals for each calendar day, rated from your usage.
- Baseten — OIDC and AWS AssumeRole for training jobs
Training jobs can now use OIDC or AWS AssumeRole during setup to pull private container images and download model weights or training data without storing long-lived…
- OpenHands — v1.18.0
Added full-size image attachments, cloud automation editing, automation run identity display, and DB-driven free/default model flags.
- Krisp — Web updates for August 31 – September 6: Call Center AI
This batch of updates improves deepfake risk visibility and gives Admins more precise Voice Translation tracking and controls.
- ElevenLabs — Scribe v2 Medical
Scribe v2 Medical is now generally available for batch speech recognition specialized for medical and clinical audio.
- Tonic Textual — v489
Fixed white gaps in V1-redacted PDFs.
- Tonic Textual — v488
Fixed dataset-by-name API requests to return HTTP 404 for missing datasets and improved Python SDK dataset listing reliability during concurrent deletions.
- Ghostwriter — Next gen multimodal agents: an interface that morphs with the conversation
Sierra’s multimodal agents bring voice, text, and visuals into the same conversation.
- Together AI — Introducing preemptible compute: the same compute, half the price
Together GPU Clusters added preemptible compute at 50% of the on-demand rate, with a five-minute drain window.
- Vyond Studio — September 2026 Vyond Newsletter
Vyond Turbo can create videos with animated or realistic narrators and automatically generated background music.
- Birdeye Reviews — Meet Front Desk agent: Your AI-powered front desk
Front Desk agents handle conversations across Voice Call, Web Chat, Text, and Email, answer questions, and manage appointments.
- You.com — The n8n You.com Node Now Has 7 Operations, Including Background Research
The n8n You.com Node now has 7 operations, including Background Research.
- Mistral Le Chat — User and Permissions Management in the Admin Panel
Manage users, groups, and roles; assign members to workspaces and create groups to bulk-assign permissions.
- Swapcard — Email Domain Manager: Set Up and Verify Your Sender Domains Yourself 📬
Organizers can set up and verify sender domains in Studio's Email Domain Manager.
- Kagi Assistant — One login for all of Kagi, and Translate reopens to everyone
Mobile apps can share images or files into new or existing threads; thread retention, sorting, and message reporting are available.
- Kagi Search — One login for all of Kagi, and Translate reopens to everyone
Kagi now has a single account for all products; Translate is freemium, and its browser extension adds read aloud and auto-proofread.
- Birdeye Insights AI — Meet Front Desk agent: Your AI-powered front desk
Front Desk agents handle customer conversations across Voice Call, Web Chat, Text, and Email.
- PromptLayer — Deployment 1
Structured Search accepts Ctrl+Enter; FOR_LOOP column variable mappings are preserved; multi-value badges handle overflow and text wrapping.
- Speak — Live tutor lessons, powered by OpenAI's GPT-Live-1
Tutor lessons are powered by OpenAI's GPT-Live-1.
- Taskade — v6.246.0 — Trusted App Restores & Remembered AI Models
Restore earlier app versions with publishing-state handling, and Taskade EVE remembers your model choice across sessions.
- Modal — Synchronized time cursor across charts
Hovering a timeseries chart now highlights the same moment in every other chart — across tabs and windows.
- IBM watsonx — Removal of fine tuning
Fine tuning of foundation models is removed. Fine Tuning - LoRA and QLoRA APIs, Fine Tuning SDK, and Tuning Studio are no longer available.
- IBM watsonx — New environments for Data Refinery
The new Default Spark 3.5 & R 4.3 environment and Default Data Refinery with Spark 3.5 & R 4.3 are added.
- IBM watsonx — Deprecation and removal of Decision Optimization runtime do_20.1
Decision Optimization runtime do_20.1 is deprecated; you can no longer use it to create new assets. It will be removed on 13 October 2026.
- Mubert Render — Meet Cast: Make Your Podcast Sound Finished
Upload your recording and edit it through the transcript; remove filler words, shorten long pauses, clean up the voice, bleep explicit language, and generate intros…
- Spellbook — Introducing Autonomous Contract Management
Spellbook introduced Autonomous Contract Management.
- Spellbook — Introducing Spellbook for Google Docs
Spellbook introduced Spellbook for Google Docs.
- Spellbook — Introducing Proofread