AI news
This page is the database. Every AI note I track lives here: model launches, coding agent changes, and the date they shipped. The homepage lists the last five titles. My own product updates stay on News.
I add items by hand. I link the vendor post or the repo when I have one. This started on September 19, 2026.
Want new items in your feed reader? Subscribe to the AI news RSS feed.
Google released Gemini 4 Neon
Gemini 4 Neon is built for complex coding, analytical reasoning, and enterprise work. It has a one-million-token context window, with API pricing of $3 per million input tokens, $0.15 for cached input, and $15 for output.
The rollout includes the paid Gemini API tier, Gemini for business, and Google AI Pro and Ultra subscribers in the Gemini app and Google Antigravity. GitHub also made Neon generally available in Copilot for Pro+, Max, Business, and Enterprise users, with a gradual rollout across its clients.
Google rolled out Genie 4 in Project Genie
Eligible Google AI Ultra subscribers can now use Genie 4 to create interactive worlds, edit environments with prompts while exploring, and generate native audio that changes with the scene. Google AI Pro members can join the waitlist.
Project Genie also gained an agent for creating worlds through conversation and a community gallery for sharing and remixing them. Genie 3 remains available, with sessions extended from 60 seconds to three minutes per generation, linked worlds, and support for 360-degree panorama inputs.
Google brought Gemini to lower-memory Android phones
The Gemini app now supports eligible Android Go devices. Google rebuilt the app to run on phones with less memory, including Gemini Live conversations in more than 45 languages, camera-based questions, and Nano Banana image generation.
You can download it from Google Play or launch it with the power button on supported phones. Google says upcoming devices from partners such as Transsion will ship with Gemini preloaded.
GitHub Copilot added eight service connectors
Copilot added GitHub-verified MCP connectors for Box, Buildkite, Calendly, ClickUp, Mixpanel, Supabase, Webflow, and Wix. Connect a service once to use it across Copilot, including the app, CLI, VS Code, and github.com.
Connectors are in public preview for all Copilot plans. Business and Enterprise admins must enable Editor preview features, and may need to update their MCP allowlist. Some connected services also require approval from their own admin.
GitHub made Copilot team plugin defaults generally available
Copilot Business and Enterprise teams can now share a default set of Agent Plugins. Team maintainers and organization owners choose approved plugins in the team’s Customize Copilot tab on github.com.
The defaults apply to direct team members in VS Code, the Copilot app, and Copilot CLI. Child teams do not inherit them. Members can turn off optional plugins where policy permits, and organization and enterprise policies still take precedence.
Gemini expanded task handling to free users
Google is rolling out Gemini tasks free of charge in the US, Japan, Korea, Canada, Australia, and India, powered by Gemini 4 Hydrogen. Switch from Chat to Task to hand off work such as scheduling appointments, comparing purchases, or analyzing bills.
Advanced features start in the US, including phone calls, Google Password Manager and Google Pay integrations, texting Gemini, and forwarding emails. Google says Gemini is designed to ask for confirmation or pre-approval before purchases. You can also set a spending cap. Calling is unavailable in Indiana, Louisiana, Minnesota, Montana, and Nebraska.
Google Search AI Mode added temporary threads
Temporary threads let you ask AI Mode questions without saving them to your history. They do not use past preferences or affect future personalization, and disappear when you close them.
Google is rolling the feature out globally on desktop and mobile browsers, across all languages supported by AI Mode. Open a temporary thread using the icon in the upper-right corner.
Google Vids made personal avatars free for personal accounts
Anyone with a personal Google account can now create a personal avatar in Google Vids at no cost. Scan a QR code to capture your likeness, then type what you want the avatar to say.
Avatars are restricted to the account holder’s likeness and stay saved in the Vids project for reuse across scenes. You can try the feature in Vids on desktop.
Microsoft announced new MAI coding, image, voice, and security models
Microsoft introduced MAI Code 2 and Code 2 Flash for everyday coding, alongside Cyber 2 Flash, Image 2.7 and Image 2.7 Flash, Realtime 1, and Dragon Copilot Transcribe built on MAI Transcribe 2. Code 2 and Code 2 Flash are available in GitHub Copilot; Code 2 Flash also has a limited-access Foundry preview.
Both image models are available in Foundry. Realtime 1 starts at $0.01 per audio minute and can be tried in MAI Labs. Microsoft also announced Licensed Weights, giving eligible customers access to MAI models they can customize and run in their own Azure environment.
Microsoft-Decision-1.1 became generally available
Microsoft-Decision-1.1 is available in Microsoft Foundry, OpenRouter, and Vercel for routing, classification, prioritization, and workflow control. It now accepts images, including photos and screenshots, and returns confidence scores so applications can defer uncertain decisions.
Foundry Global Standard pricing is $0.042 per million input tokens, with output tokens free. Microsoft reports more than twice the speed of its first decision model at the same cost.
Microsoft shipped Copilot Home and Office, with Autopilot entering rollout
Copilot Home, which brings Chat and Cowork together, and Office inside Copilot are now generally available. Microsoft also released Copilot Managed Runtime for hosting and managing apps in an organization’s tenant. Copilot Code is due to become generally available in the coming weeks.
Autopilot is starting to roll out to Frontier customers. It has its own identity, memory, and tools for ongoing work in Teams, Outlook, and documents. Copilot Workspaces, a shared space for people and agents, is also available in Frontier.
OpenAI introduced ChatGPT for Sales
ChatGPT for Sales adds a Sales plugin with workflows for account research, call preparation, proposals, and follow-ups. OpenAI also announced more than 30 new or expanded partner plugins, including connections to CRM systems, prospecting tools, and meeting notes.
Business and Enterprise admins can enable the Sales plugin and relevant partner plugins for their workspace. Availability depends on the plan, region, and workspace settings; members connect the accounts their workflows need.
OpenAI updated Astra for Law and expanded access requests
Astra for Law now uses GPT-6.2 Astra with an expanded legal search index and instructions for legal research and writing. All ChatGPT Enterprise customers can request access through the Trusted Access Program, including API access.
OpenAI reports that the configuration earned full credit on 20.8% of Harvey LAB-AA assignments, compared with 15.0% for standard GPT-6.2 Astra. Its legal search index has grown by 34% since September to more than 310 million URLs.
OpenAI released GPT-6.2 Astra and Sol
OpenAI released GPT-6.2 Astra for its hardest professional tasks and GPT-6.2 Sol for everyday work. Both are available in ChatGPT Work and Codex on Plus, Pro, Enterprise, and Edu plans, with admin enablement for Enterprise and Edu. Developers can also use both models through the API.
OpenAI reports better coding, computer use, and document creation with fewer output tokens. Astra's Standard API rates fall to $8 per million input tokens, $0.40 for cached input, and $40 for output. Sol keeps the previous Sol rates of $2, $0.10, and $10 respectively.
There's also an integration change for developers: GPT-6.2 can drop earlier reasoning after edits to conversation history. OpenAI recommends preserving the existing prefix and appending new instructions.
Gemini Notebook can turn sources into interactive sites
Gemini Notebook can now build interactive sites from your sources. Ask for a visual explainer, a dashboard, or a project tracker in chat, then open the result in a new tab or share it.
Google is rolling this out globally on the web for users aged 18 and older, including Workspace subscribers. Mobile support is coming later.
Gemini is separating chat history from model training
Google announced separate controls for saving Gemini chat history and using it to train generative AI models. You will be able to keep your history while opting out of training, and still choose how long saved history stays.
The changes roll out globally in the coming weeks. Turning History off will still leave future chats stored for around two months for responses, feedback processing, and safety.
Gemini added 66 regional voices
Google is rolling out 66 regional voices in Gemini on Android and iOS. Options cover US, UK, and Indian English, Hindi, Brazilian Portuguese, Indonesian, and Korean.
Your chosen voice applies across Gemini Live and spoken chat responses. You can preview the options in Gemini Live or the voice settings.
Google released AdMob skills for coding agents
AdMob agent skills give coding agents playbooks for migrating to the Google Mobile Ads Next-Gen SDK, checking SDK integrations, and adding ad formats. The skills are available now.
Google also announced Ask AdMob, a Gemini assistant for monetization questions, and an MCP server for account data and ad setup changes. Ask AdMob enters testing over the next few months; the MCP server is in early testing, with broader developer access planned for early 2027.
Ask Maps added trip planning and expanded across Europe
Ask Maps is adding trip itineraries and search across saved lists. Google also announced route tips and appointment finding for beauty, pet care, auto, and home repair services.
The rollout starts on Android and iOS. Itineraries and list search work wherever Ask Maps is available, including its expansion to 30 European countries. Route tips and appointment finding launch first in the US.
OpenAI is expanding free ChatGPT access for researchers
OpenAI is expanding ChatGPT for Academic Researchers toward free access for up to 200,000 researchers worldwide. The next application round opens November 23, with 50,000 new grants awarded to eligible applicants on a first-come, first-served basis.
Each grant covers ChatGPT Plus for the recipient and up to two collaborators. Participants also receive training, and data shared through the program is excluded from model training by default.
Alibaba released Qwen4-Max with native vision and a million-token context
Qwen4-Max starts the new Qwen4 model family with native vision and a one-million-token context window. Its backbone has 2.4 trillion parameters, with 71 billion active per token. You can use it through the QwenCloud Token Plan or API, and in QwenWork, Qoder, and QwenStudio.
The team also open-sourced the Qwen Live Delivery plugin. It lets long-running agents deliver an early usable version, then improve it while you review results and change direction. Smaller Qwen4 models are planned for open source; the announcement does not release Max’s weights.
Cloudflare open-sourced Kitesurf
Kitesurf, Cloudflare’s browser engine for AI agents, is now on GitHub under the Apache 2.0 license. You can read the code, run the engine on your computer, and deploy your own copy to Cloudflare Workers instead of using the managed Browser Run API.
Running it on Cloudflare needs the Workers Paid plan, because Kitesurf starts a Dynamic Worker for each page. The engine has no authentication of its own, so put Cloudflare Access in front of it as soon as you deploy.
Cloudflare announcement →Kitesurf on GitHub →A deep dive into Kitesurf →
Google AI Studio added Cloud credit redemption and billing setup
Google AI Pro and Ultra subscribers can now claim their monthly Google Cloud credits directly inside AI Studio. The same flow creates a Google Developer Program profile if needed, links a billing account, and lets you track credits and Gemini API usage.
Each monthly grant lasts 12 months, and unused credits roll over. Only the primary subscriber can use them. Gemini API billing requires a minimum $5 prepayment and balance before promotional credits can be used.
Gemini added Google Wallet for personal finance questions
Google is rolling out a Google Wallet connection to Gemini app users in the US, on web, Android, and iOS. You can ask about spending, find recurring subscriptions, and build budgets or charts from your transactions. Linking bank and credit card accounts through Plaid adds more information.
The connection is optional and read-only: Gemini can read account information, but cannot move or invest money. Google also added Memory Summary, where you can review and edit saved preferences and goals, wherever Personal Intelligence is available.
OpenAI started rolling out ChatGPT Wallet
ChatGPT Wallet lets a dot complete purchases using a payment token instead of your actual card details. Before each purchase, you review the items, total, and selected card, then approve with a security check such as Face ID or a fingerprint scan.
OpenAI built it with American Express, Mastercard, and Visa. The rollout starts in dots for ChatGPT Pro subscribers aged 18 and older in the US with eligible US-issued cards. OpenAI also published a Commerce Partner Playbook to help businesses support agent purchases through their existing websites and checkout systems.
DeepSeek opened its application APIs to developers
Developers can now use DeepSeek’s web search, file input, URL reading, and safety checks through its existing APIs. With deepseek-flash, these features work in the Messages, Responses, and Chat Completions APIs. The service handles tool calls and assembles the context.
Search, page fetching, OCR, speech recognition, and text-to-speech are also available as standalone APIs. DeepSeek released an open-source demo chat client that can import your DeepSeek chat history.
DeepMind opened the datasets behind its AI weather models
Google DeepMind released live feeds and historical archives for SatWeave and PARDIG, two datasets used by WeatherNext 3. SatWeave combines satellite imagery into an hourly global mosaic. PARDIG uses AI to estimate precipitation at roughly 10-kilometer resolution in 15-minute intervals, with an average delay of about five hours.
Both datasets are openly available on Google Cloud Storage. PARDIG’s historical catalogs are also available in Google Earth Engine. Researchers and developers can use them for weather, hydrology, and climate applications.
Anthropic launched Claude for Manufacturing
Claude for Manufacturing is a plugin for Team and Enterprise plans with 17 skills for quoting jobs, investigating failed inspections, diagnosing breakdowns, comparing suppliers, and reporting on production.
It connects to Autodesk Fusion, Infor, Microsoft 365, and Oracle NetSuite. Teams can also give Claude exports from systems without a connector through Microsoft 365. The skills prepare work for a person to review and approve. A separate Infor plugin can change records after approval, using the person’s existing permissions.
Claude Meetings started rolling out in beta
Claude can now join a video call as a visible participant. You invite it like anyone else, or add it to an upcoming meeting from your calendar in claude.ai. People ask it questions out loud or in the chat, and it works on docs, whiteboards and other artifacts while the call goes on, then writes the notes and action items.
Each meeting gets a page in claude.ai with the notes, the transcript and everything Claude built. Claude keeps a transcript, a summary and descriptions of shared screens, not the audio. By default it stays out of calls with people from outside your organization, and leaves if one joins partway through, unless an admin allows it.
It’s rolling out on Team and Enterprise plans, and only works in Google Meet for now. On Team plans it’s on by default, while on Enterprise an admin turns it on. Organizations with zero data retention, a HIPAA-ready setup or customer-managed encryption keys can’t use it yet. During the beta, notes, transcripts and answers are free, except answers that use a connected tool, and artifacts made in a call are free until February 10, 2027. Anthropic announced Meetings as coming soon on November 10, together with the Claude Sheets and Whiteboards betas.
Conductor will keep being developed after all
Eight days after announcing that Meta acquired Conductor, Charlie Holtz said the team decided to continue development of the app, joking that reports of its death were greatly exaggerated by the team itself. Regular updates resume next week.
The plan announced on November 4 was a wind-down: Conductor Cloud read-only on November 13 and shut down on November 20, and the local Mac app getting only new models and critical fixes until early 2027. Holtz’s post doesn’t say what changes for Conductor Cloud or for the Meta deal, so if you rely on the cloud, keep a local copy of your workspaces until Conductor says otherwise.
Claude Managed Agents can now be self-hosted, in private beta
Until now you could run only the sandbox of a Claude Managed Agent on your own machines, while Anthropic ran the rest. The self-hosted version moves the whole service into your cloud or data center: the server that runs the agent loop, its PostgreSQL database with the session history, and the workers that run the agent’s shell and file tools, each session in its own sandbox container.
Claude itself stays outside, and you pick where model requests go: the Claude API, Amazon Bedrock, Google’s Agent Platform, Microsoft Foundry, or your own gateway. Credentials for MCP tools and supported API calls stay out of the sandbox, in your own secrets manager if you want, and every environment can be limited to a list of allowed hosts. Each tool call and its result is logged and can be exported with OpenTelemetry.
The docs cover a Docker Compose quickstart and a Kubernetes deployment (1.30 or later). You keep the server, the database and the firewall rules patched yourself. Access goes through your Anthropic account team, and Anthropic hasn’t published pricing.
Google released Gemini 4 Hydrogen
Gemini 4 Hydrogen is the fast model of the Gemini 4 family, next to Carbon and Argon, built for everyday coding and agent work. Google presents it as the upgrade from Gemini 3.8 Flash: it finishes tasks 2.5 times faster and solves 57% of Terminal-Bench 4.1, up from 18.2%.
It has four thinking levels: low, medium, high (the default), and a new xhigh. Google reports 51 on the Artificial Analysis Intelligence Index at high, one point above Argon, and 54 at xhigh. Antigravity also gets Orchestrator, which switches between Carbon and Hydrogen inside one session. On Terminal-Bench 4.0, Google says it came close to Carbon’s success rate using 24% less quota.
It’s rolling out today in the Gemini API and Google AI Studio as gemini-4-hydrogen, free tier included, plus the Gemini app, Search, Gemini for business, and Antigravity for Google AI Pro and Ultra subscribers. It costs $1.50 per million input tokens and $7.50 per million output tokens, flat across the 1 million token context window. That’s twice the introductory price of Gemini 3.8 Flash, and the same as its standard price from January 1, 2027.
Google launched the Gemini Decisions API
The Gemini Decisions API uses the new gemini-decision-1 model to score predefined choices. It handles yes/no checks, multiple choice, and rubric scores across text, images, audio, video, and PDFs, with up to 1,000 questions sharing one input.
Each answer includes confidence scores, useful for tool selection, model routing, and policy checks. Google reports 150 to 250 milliseconds of end-to-end latency for contexts up to 16K tokens. The API accepts up to 256K input tokens and costs $0.10 per million input tokens, with no output tokens billed. It is available through the Gemini API in Google AI Studio.
DeepMind introduced AlphaGenome 2
AlphaGenome 2 predicts how DNA variations affect specific cell types and states. Researchers give it a DNA sequence and a biological condition, so it can examine effects that broad tissue averages could miss.
DeepMind says the model identifies nearly twice as many variants that change gene expression as its previous model. It trained on more than 100 million cell profiles across 68 tissues. Access is available through the AlphaGenome API for non-commercial research, with commercial research in private preview.
MiniMax released H3-Image with open weights
MiniMax released H3-Image for generating and editing images, including generation from reference images. It supports up to 16 references. MiniMax says it can generate a 4096 by 4096 image in under 10 seconds.
The weights are available on Hugging Face. You can also use H3-Image through the MiniMax API Platform, MiniMax Code, and Hailuo AI.
Grok Bot 1.0 shipped with meetings and Microsoft Teams
Grok Bot is out of beta. SpaceXAI announced version 1.0 on November 2, and the app release that carries the number, 1.0.1, landed on November 11.
A Bot can now join a Google Meet, Zoom, or Microsoft Teams call, answer in the meeting chat, and send a summary afterward. Meetings are on the Team, Enterprise, and Ultra plans.
Team Bots can be shared with specific people or groups and brought into Microsoft Teams as well as Slack. What teammates teach a Team Bot now waits for the owner or a Manager to approve it, and owners can share an app login with the team once an admin allows it.
Grok Bot can join your Tailscale network
A Bot can now join a personal or company tailnet on Tailscale and reach the private MCP servers, internal tools, and apps on it. You connect it from the Tailscale plugin, and adding an MCP server on a private network asks for your approval first.
The Bot reaches only what your Tailscale access rules let its device reach, and its web traffic still goes out through the cloud. All your Bots share the computer that joins.
Matt Palmer pointed out the setup this opens up: run an MCP server on a Mac mini or a Raspberry Pi at home, put it on your tailnet, and the Bot can use it from anywhere, without deploying the server publicly. His examples are a Plex library, custom coding agent workflows, and local LLMs.
Grok Bot announcement →Matt Palmer on X →Join your Tailscale network →Tailscale guide →
Claude added Sheets, Whiteboards, and full apps inside chat
Claude Sheets lets teams build spreadsheets with Claude, edit formulas, and work across tabs. Claude Whiteboards adds a shared space for organizing ideas. Claude Data, formerly Dashboards, connects to tools such as Snowflake and Salesforce to build dashboards and visual reports.
Supported interactive connectors can now open a full app beside the conversation or expand it to full screen. You and Claude can edit in the same tool, including Confluence, Notion, Adobe, and Box. The update is rolling out on all plans on the web and desktop as accounts receive the new Claude experience or organizations enable it.
Developers can build these interfaces with MCP Apps. Support for app tools and embedded sites is in public preview.
Sheets and Whiteboards announcement →Interactive connectors announcement →
Cursor Projects added multiplayer collaboration
You can now invite teammates into a Cursor Project and work alongside their agents. A shared task board shows who is doing what, what agents are working on, and the pull request behind each task.
Each person keeps their own chats with their coordinator and worker agents. Shared documents, requirements, and decisions give the whole Project common context, and agents can communicate across teammates to coordinate overlapping work. Multiplayer Projects is rolling out in beta to all Cursor users.
Google opened Gemini 4 Carbon to developers and Ultra subscribers
Gemini 4 Carbon is now available to paying developers and enterprises through Google AI Studio and Gemini for business. It is also rolling out to Google AI Ultra subscribers in Antigravity, the Gemini app, and AI Mode in Search.
Introductory API pricing is $2 per million input tokens and $10 per million output tokens. Google positions Carbon for complex coding, debugging, and scientific research, while recommending Argon for most software engineering and knowledge work.
Google is also previewing /swarm in Antigravity, which turns a prompt into a coordinated team of agents. Carbon can handle planning and review while faster models work on parallel tasks.
Google introduced Gemini Robotics 2.5
Gemini Robotics 2.5 is Google DeepMind’s new vision-language-action model, rolling out to its robotics partners. Google evaluated it across more than 200 tasks, including handling delicate objects and using tools.
Google reports a 96% success rate for picking and packing automotive parts with Apptronik’s Apollo 2 humanoid in a Mercedes-Benz production-line proof of concept. Across its evaluations, it reports tasks completing 20% to 30% faster than previous baselines.
DeepMind introduced AlphaProof Code for software verification
Google DeepMind introduced AlphaProof Code, a framework that combines human experts with AI agents to prove that software meets mathematical specifications. The team used it to verify key security properties of Signal’s core cryptographic implementation.
Human experts reviewed the specifications, then Gemini Carbon agents generated more than six million lines of Lean proofs. Lean checked the proofs automatically. DeepMind says the process took four weeks and is open-sourcing the verification artifacts.
OpenAI launched Businesses in ChatGPT in beta
Businesses can now claim their organization in ChatGPT through Business Hub. After verification, they can update business details, supply product feeds with pricing and availability, and see aggregated mentions, website clicks, and discovery topics.
Eligible ecommerce and online service businesses in the United States can also build and launch a Business Agent to answer questions about their offerings. These conversations are clearly labeled and separate from ChatGPT’s own answers. OpenAI says businesses receive aggregated insights rather than users’ conversations.
Meta released Muse Ember, its smallest Muse model
Muse Ember is Meta’s smallest and most efficient model, built for repeated agent work like tool calls, coding subtasks, routing, extraction, and parsing long documents. Meta says it’s the first model trained using Muse Wonder Full.
It takes text, images, audio, and video, with a 1 million token context window and up to 512K output tokens. Meta compares it with Claude Haiku 5.5 and GPT-6 Luna: it scores 40.9% on Terminal-Bench 4.0 against 39.2% for Haiku, and 37.8% on AutomationBench-AA against 13.9%. Haiku still leads on FrontierCode 2.0-AA, 34.9% to 27.2%.
Pricing is $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 200K tokens, and $0.30 and $1.50 above that. It’s available as muse-ember-1.0 through the Meta Model API, in Muse Code, and on Google Cloud and Oracle Cloud.
Anthropic released Claude Fable 5.2 and Mythos 5.2
Anthropic upgraded its models for ambitious technical work, including coding, reasoning, and vision. Fable and Mythos share the same underlying model. Fable has stronger cybersecurity and biology safeguards; Mythos is available to vetted organizations through verification programs.
Anthropic says they generate output up to 30% faster than the 5.1 models. Standard API pricing is $8 per million input tokens and $40 per million output tokens. An optional fast mode costs twice as much and speeds up token generation by up to 2.5 times.
Fable 5.2 is available on the Claude Platform as claude-fable-5-2, and through AWS, Google Cloud, and Microsoft Azure. Anthropic still recommends starting with Opus 5.5 for most everyday coding and knowledge work.
OpenAI introduced GPT-6 Cyber
GPT-6 Cyber is OpenAI’s new model for finding and validating exploitable software vulnerabilities. OpenAI says it improves on GPT-5.6 Cyber while following the boundaries set by security researchers more reliably.
OpenAI reports 48.5% on ExploitGym, up from 32.1% for GPT-5.6 Cyber. Its researchers also used the model to identify 81 previously unknown high-severity Firefox vulnerabilities and develop candidate patches. Mozilla has released fixes for many of them.
Access starts through the API and Codex for selected approved Daybreak Red customers. OpenAI plans to expand it to all Daybreak Red customers, and classifies the model as Critical in cybersecurity under its Preparedness Framework.
Mistral expanded Studio with managed agents and apps
Mistral expanded its enterprise Studio platform with managed agents and apps, plus updates to workflows, enterprise context, observability, and evaluations. Teams can deploy in the cloud, in their own VPC, on-premises, or across a combination of environments.
Agents can use skills, code sandboxes, subagents, and hooks for custom logic. Studio manages their runtime and lets teams inspect their work. A new Mistral MCP server lets developers build with Mistral Vibe or another coding agent, test locally, and deploy to Studio with one command.
Google released Gemini Omni 1.2 Flash for longer videos
Gemini Omni 1.2 Flash generates videos up to 15 seconds long, up from 10 seconds in Omni 1.1. It adds multi-turn conversational editing and improved reference following, so you can adjust a clip without starting again.
It is generally available through the Gemini API and Gemini for business. Google cut the price by 30% from Omni 1.1, to $0.07 per second for 720p video. You can draft at 360p or 720p and upscale selected clips to 1080p or 4K.
Moonshot releases the Kimi K3.1 Pro open weights
Moonshot AI published the full weights of Kimi K3.1 Pro on Hugging Face, on the November 4 date it promised when the model launched in its apps and API on October 23.
It’s a 2.8 trillion parameter mixture-of-experts model with 104 billion active parameters, a 1 million token context window, and text and image input. The repo is 96 safetensors shards, about 1.56 TB, shipped in MXFP4 because Moonshot trained it that way. vLLM and SGLang support it from day one.
The Kimi K3.1 Pro license lets you use, modify, fine-tune and redistribute it for free. Companies selling it as an API with more than $20 million in yearly revenue need a separate deal with Moonshot, and very large products have to show the Kimi K3.1 Pro name.
Meta acquired Conductor, the parallel coding agent app
Meta acquired Conductor, the Mac app that lets an engineer run a team of coding agents in parallel, from the AI coding app at conductor.build. Terms weren’t disclosed. Conductor was founded in 2024 by Charlie Holtz, who was CEO, and Jackson de Campos, who was CTO.
Holtz confirmed the deal himself, saying Conductor is joining Meta and that the team is excited to work with Nat Friedman and Alexandr Wang at Meta Superintelligence Labs. Meta reportedly wants Conductor’s pace and attention to developer experience for its next generation of AI coding products.
The deal came the same day OpenAI told Axios it plans to acquire Pierre Computer, which stores and manages AI-generated code, after its plan in June to acquire Ona, which works on cloud coding environments and agent orchestration. The big labs are buying up the developer tools layer, not just shipping models.
Mistral released the open weights of Mistral Large 4
Mistral published the weights of Mistral Large 4, its flagship model, which was in preview through the API earlier this month. This is an improved version of that preview, with more reinforcement learning on coding and defensive cybersecurity, and better reasoning efficiency so it costs less to run.
It’s a mixture-of-experts model with about 1 trillion total parameters, 49 billion active per token, a 1 million token context window, and text and image input. Mistral says it’s the best open-weight model for coding. You can download FP8 and NVFP4 checkpoints from Hugging Face and run them with vLLM, SGLang, Transformers, or llama.cpp.
It’s also on the Mistral API as mistral-large-4 and on OpenRouter, at $0.70 per million input tokens and $2.10 per million output tokens. The license is a modified MIT that allows commercial use and fine-tuning, but companies selling it as a hosted model service need a separate license from Mistral.
Google announced Gemini 4 Carbon
Gemini 4 Carbon is Google’s new frontier model, building on Argon. Google reports stronger coding and long-running agent performance, and a score of 59 at the top of Artificial Analysis’s Intelligence Index v5.0.
Cybersecurity is a major focus: Google reports 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71% on CWE-bench for patching. These are Google’s reported results, evaluated without guardrails.
Access starts this week with selected cyber defenders through the Fairwind Program. Wider access for developers, businesses, and consumers is still to come. Google announced introductory API pricing of $2 per million input tokens and $10 per million output tokens, with cached input at 95% off the input price.
OpenAI released GPT-6.1 Terra
GPT-6.1 Terra is OpenAI’s new fast, cheaper model in the GPT-6 family, replacing GPT-5.6 Terra. It’s built for everyday work like writing, visual reasoning, and computer use, and OpenAI says it finishes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.
It’s also a big jump in coding over the old Terra: on Terminal-Bench 4.0 it solves 50.3% of tasks at max reasoning, up from 21.5%. OpenAI still points to GPT-6.1 Sol and GPT-6 Astra for the hardest coding work.
It’s available today to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.1-terra at $1 per million input tokens and $5 per million output tokens, half the price of GPT-5.6 Terra.
Moonshot releases Kimi K3.1 Pro
Moonshot AI released Kimi K3.1 Pro, the first model in its new K3.1 series. It’s a post-trained version of Kimi K3 with a lot more post-training compute, a 1 million token context window, and Moonshot calls it its most capable open-weight model so far. There’s also a K3.1 Pro Highspeed variant that runs 2 to 3 times faster.
You can use it today on Kimi.com, in the Kimi apps, in Kimi Code, and through the API as kimi-k3.1-pro, priced at $2 per million input tokens and $10 per million output tokens. Moonshot says it improves mostly on coding, agentic tasks, financial research and 3D/CAD work, and it reports 65.7% on Terminal-Bench 4.0, right behind Claude Opus 5.5.
The weights aren’t downloadable yet. Moonshot says the full model weights will be released by November 4.
Reflection released Beam, an Apache 2.0 open-weight model
Beam is the first model from Reflection, a new US lab, and the weights are now on Hugging Face under the Apache 2.0 license. It’s a mixture-of-experts model with 501 billion total parameters, 23 billion active per token, and a million-token context window, built for coding, reasoning, and long-running agent work.
Reflection ships the weights in BF16, FP8, and NVFP4, with day-one support in Transformers, vLLM, SGLang, and Unsloth. You can also use it through Reflection’s own API or hosted providers like OpenRouter, Together AI, Fireworks AI, Cloudflare, and Ollama.
Alongside the model, Reflection open-sourced Mirror CLI, a terminal coding agent trained together with Beam. Beam also works with other harnesses such as Pi, OpenCode, and Hermes.
Meta released Muse Spark Open with downloadable weights
Muse Spark Open is Meta’s new open-weight model for coding and agents. It accepts text, images, video, and audio, and produces text. It has 816 billion total parameters, 42 billion active per token, and a million-token context window.
Running it yourself takes serious hardware. Meta lists at least 700 GB of GPU memory for the NVFP4 version, and demonstrated local agent workflows on two M5 Ultra Mac Studios with 512 GB each. Hosted access is available through providers including Fireworks AI, Groq, OpenRouter, and Together AI.
The weights use the Muse Model License, which allows many commercial uses but includes attribution and usage conditions. Model-as-a-Service businesses above its $20 million aggregate revenue threshold need a separate agreement with Meta.
Google made Gemini 4 Argon available to everyone
Gemini 4 Argon, Google’s new frontier model, is now available to everyone, not just the cyber defenders who got it first in September. Google says it ranks first on the Vals Index, which covers coding, finance, legal, and tax work.
Developers can use it through the paid Gemini API in AI Studio, and Google AI Pro and Ultra subscribers get it in the Gemini app and in AI Mode in Search. Ultra subscribers can also use it in Google Antigravity, which is now built into the Gemini macOS app.
Introductory API pricing stays at $2 per million input tokens and $10 per million output tokens until November 30, then it goes to $4 and $20. The API handles up to 1M tokens of input and output combined.
Anomaly released OpenCode 2
OpenCode 2 is a new version of Anomaly’s open source coding agent. The terminal, the desktop app, and the web app now talk to one background program on your computer, so a session you start in one place shows up in the others.
The model can write a short script that calls several tools in one step, instead of calling them one at a time. It can also hand work to a second agent, or start a long command, and keep going while that runs.
Plugins and anything built on the old server API need a rewrite. Settings, agents, and project files from OpenCode 1 are meant to keep working. The program is free. You pay for the model, or you can use Anomaly’s hosted models.
Meta released Muse Wonder
Muse Wonder is Meta’s new frontier model, and the step after Muse Spark. It comes in two versions, Muse Wonder and the bigger Muse Wonder Full, and Meta says both lead on personal and professional agent work, memory, and coding.
Meta highlights two abilities. Active Memory keeps track of what matters to you across conversations. Active Focus keeps the model working on a hard goal, trying other approaches and checking in before important actions.
Muse Wonder is available today through the Meta Model API and Muse Code. Muse Wonder Full is for Muse Code subscribers. It will power the Muse app soon, starting with subscribers.
OpenAI brought GPT-6 to everyone in ChatGPT, with Intelligent UI
GPT-6 is now the model behind the Chat tab in ChatGPT for everyone, free users included. Paid plans get GPT-6 Sol, while Free and Go get GPT-6 Luna. They replace GPT-5.6 Sol and GPT-5.6 Luna.
The big new thing is Intelligent UI. Answers aren’t just text anymore: they can include charts, interactive diagrams, buttons, forms, and small tools like a calculator or a bill splitter you use right in the conversation. GPT-6 can also start answering while it’s still thinking.
It started rolling out to Plus, Pro, Business, and Enterprise on October 7, and to Free and Go on October 8. The models used in ChatGPT Work and Codex don’t change with this release.
Anthropic released Claude Haiku 5.5
Claude Haiku 5.5 completes the Claude 5.5 family. Anthropic calls it its cheapest, fastest, and most capable small model, built for high-volume work like summaries, classification, browser use, and subagents working under Opus 5.5 or Sonnet 5.5.
It costs around 75% less to run than Haiku 4.5. For prompts up to 100k tokens it is $0.10 per million input tokens and $0.50 per million output tokens. It is also the first Haiku with an adjustable effort setting.
It is available today on the Claude Platform as claude-haiku-5-5, and on AWS, Google Cloud, and Microsoft Azure. Anthropic also cut Sonnet 5.5 cache-read prices in half, and is adding a monthly API credit for Max and Team subscribers.
Mistral released Mistral Large 4 in preview
Mistral Large 4 is Mistral’s new flagship model, and its largest so far. It has 1 trillion parameters, 49 billion of them active for each token, and it understands both text and images.
Mistral says it is the strongest open-weight model built outside China, with strong results on coding, agent work, cybersecurity, finance, and law. In a blind coding evaluation Mistral ran, it placed second of five models, behind only Claude Opus 5.
The preview is available today through the Mistral API in Mistral Studio. The weights come out by the end of October, so anyone will be able to download and run it.
Earendil released Pi 1.0
Pi is a small, open source coding agent that runs in the terminal. It works with models from almost every provider, with a ChatGPT plan, or with a model running on your own computer. Earendil says hundreds of thousands of people use it every week, and calls 1.0 a stable version that people and businesses can depend on.
Pi used to leave MCP out on purpose. Now it connects to MCP servers, the common way agents talk to other apps, and the model can write a short script that calls several tools in one go. Extensions can also send each request to a different model, and the terminal screen has a new look and opens full screen.
Earendil also released Pi Durable, an experimental package for building agents that run for a long time and pick up where they left off after a crash.
Cloudflare released Clef, an open decision model
Clef is Cloudflare’s first decision model, in the same family as TypeSafe’s Jev. Instead of writing text, it answers typed questions with probabilities, so software can route a ticket, block a request, or hand a case to a person.
It comes in two sizes, Clef and a faster Clef-flash, both hosted on Workers AI. The weights are open source on Hugging Face, it accepts the same requests as Jev, and unlike Jev it can also look at images. It costs more than Jev per request.
Cloudflare is also starting a service to fine-tune Clef on your own data, working hands-on with design partners first.
Google released Gemini 4 Argon
Gemini 4 Argon is Google’s new frontier model. Google DeepMind says it is strong on long-horizon software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense.
It is rolling out first to trusted cyber defenders through Google’s Fairwind Program. Wider access for paid API customers and Google AI Ultra is planned after more guardrail work. Output context goes up to 1M tokens.
Introductory API pricing is $2 per million input tokens and $10 per million output tokens, with a steep discount on cached input. After the intro period it rises to $4 and $20.
OpenAI released GPT-6.1 Sol
GPT-6.1 Sol is an upgrade to GPT-6 Sol. OpenAI says it nearly matches GPT-6 Astra on agentic coding, computer use, and professional work, at one-fifth of Astra’s standard API input and output prices.
It is available today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu. It is not in Chat yet. In the API the model id is gpt-6.1-sol, at $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens.
A faster Ultrafast option, up to 8x token generation in Codex, is coming in the next few days. Astra still leads on the hardest scientific tasks.
OpenAI launched dots
Dots are always-on agents from OpenAI, powered by GPT-6 Astra. Each dot has its own cloud computer and browser, can connect to thousands of apps, and keeps working toward your goals in the background.
You message or call your dot in ChatGPT, Slack, or Teams. It can also reach you with progress or a decision that needs approval. Conversations with the dot do not count toward ChatGPT usage limits. Codex and ChatGPT Work tasks it starts still do.
The first dot is included for Pro and Business Premium in eligible markets. Enterprise, Edu, and Healthcare can try the beta when an admin turns it on. You create it in the ChatGPT desktop app or desktop browser.
OpenAI launched the Decisions API
OpenAI launched the Decisions API for real-time decisions. You define questions with finite pre-defined answers, provide context as text or images, and get answers for classifying content, routing requests, or choosing an agent’s next action.
It is available in limited preview today, with a broad release planned in the coming days. This puts OpenAI in the same space as TypeSafe’s Jev, with a model making bounded decisions inside software instead of generating free-form chat.
OpenAI announcement →A deep dive into Jev →A deep dive into Clef →
Anthropic released Claude Sonnet 5.5
Claude Sonnet 5.5 is the second model in the Claude 5.5 family. Anthropic says it runs over 30% faster than Sonnet 5 and costs up to 30% less for most work, while scoring much higher on coding and knowledge-work benchmarks.
It is strongest at well-scoped everyday tasks, bug fixes, and polished documents, slides, and spreadsheets. Opus 5.5 stays the pick for complex open-ended work.
Sonnet 5.5 is available in Claude, Claude Code, the API, and on Amazon Bedrock, Google Cloud, and Microsoft Azure. Haiku 5.5 is still coming in the next few weeks.
OpenAI released GPT-6 Sol and GPT-6 Luna
GPT-6 Sol and GPT-6 Luna bring much of what makes GPT-6 Astra good into faster and cheaper models. Sol is meant for complex coding and agent work. Luna is the smallest and cheapest, for simple tasks you run a lot.
Both are rolling out in ChatGPT and Codex for paid plans, and in the API at half the price of the GPT-5.6 models. Free users can try Luna in the desktop app.
Anthropic released Claude Opus 5.5
Claude Opus 5.5 is the first model in the Claude 5.5 family. Anthropic says it performs close to Fable 5.1 on most work, while costing less and answering faster than Opus 5.
It is available in Claude, in Claude Code, and on the major cloud platforms. Sonnet 5.5 and Haiku 5.5 are coming in the next few weeks.
SpaceXAI released Grok 4.7
Grok 4.7 is the new SpaceXAI model for coding and longer knowledge work. It was trained on tasks that take many hours, and it knows how to work inside Grok Bot.
It costs the same as Grok 4.6, and there is a faster variant at a higher price. You can use it in Cursor, in Grok Build, and through the Grok API.
Claude Code now reads AGENTS.md
Claude Code now reads AGENTS.md when a project has no CLAUDE.md file. If the project already has a CLAUDE.md, Claude Code keeps using that one.
You can change this behavior in the settings, including loading both files at once.
This is the file I already keep in repos so Cursor and Codex start with the same rules. Claude Code catching up means I can stop maintaining a second CLAUDE.md just for that tool.
TypeSafe left stealth with Jev
TypeSafe AI came out of stealth with $40M in funding and Jev, its first public model. Jev is not a chatbot and not a coding model. You give it data and a list of questions, and it answers each one with how confident it is.
TypeSafe sees it as a small, fast, cheap piece inside ordinary software, not something you chat with. Access is through a waitlist for now.
Cursor launched Projects
Cursor launched Projects for work that lasts longer than one chat. Each Project has a coordinator that plans the work, delegates it to other agents, and keeps shared context across local and cloud sessions.
The coordinator does not write code itself. It can also watch Slack, schedules, and pull requests, then start new work when one of those events happens.
OpenAI launched the Agents API beta
OpenAI now offers the engine behind Codex as an API, so developers can build their own long-running agents on top of it. OpenAI keeps the agent session alive and recovers it when something goes wrong.
The Agents API is in public beta for all developers.
OpenAI announced GPT-Live-1
GPT-Live-1 is a voice model that can listen while it talks, so the conversation feels more natural. When it needs to think or use a tool, it hands that work to another agent and keeps talking.
Voice is billed by the minute, and the work done by the other agent is billed separately.
Grok Bot gained Cursor cloud-agent orchestration
Every Grok Bot can now create Cursor cloud agents, read their transcripts, inspect screenshots attached as proof, and send follow-up instructions. A Bot can also interrupt a run that is going in the wrong direction.
This gives Grok Bot a way to find work outside a repository, then hand the coding part to Cursor.
Meta launched Muse
Muse is a personal AI agent from Meta. You message it in the Muse app or in WhatsApp, and it can send email, book travel, and fill out forms. It keeps working after you close the app, and it asks before it sends an email or pays.
Meta runs it in its own protected cloud environment, and payments use a one-time card so your real card number stays hidden.
It is rolling out in the US, free up to a usage limit, with two paid plans for heavier use. This is the consumer agent, separate from Muse Glimmer, the open model Meta released in August.
WebMCP added consequentialHint
WebMCP tools can now mark an action as consequential. Set consequentialHint for work that is significant or hard to reverse, such as booking a flight, transferring money, or deleting data.
The annotation defaults to false and remains a hint. The browser or agent still needs proper authorization and a confirmation step before it acts.
Announcement →WebMCP specification change →A deep dive into WebMCP →
Herdr added multi-machine agent management
Herdr 0.9 brings local and remote machines into the same terminal interface. You can add an SSH-reachable machine, see its agents beside your local ones, and move between them without opening another dashboard.
Each server still owns its own sessions and terminals. Herdr combines the view, not the underlying state.
OpenAI released GPT-6 Astra
GPT-6 Astra is the new OpenAI flagship model. It started with a small group of companies, then rolled out to paid ChatGPT plans and the big cloud platforms.
OpenAI presents it as the step after GPT-5.6 Sol for coding, browsing, using a computer, and professional work. Because it is so capable at hacking tasks, the most advanced security features stay restricted.
I have not used Astra as my daily driver yet. Sol is still the model I send precise coding work to in Cursor.
Anthropic released Claude Fable 5.1
Fable 5.1 is an update to Fable 5, at the same price.
Anthropic says it is better on long agent runs and less likely to declare success when it is stuck. That matches how I use Fable: planning, review, and judgment, not the first pass at a small code change.
Grok Bot opened to more Cursor and SuperGrok plans
Ten days after launch, xAI expanded Grok Bot beyond SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium. SuperGrok Plus, Cursor Pro+, and every Cursor Teams subscriber got in, plus a limited trial for other users.
You still sign in with a Cursor account. The Bot still runs on one cloud computer shared by all your Bots.
Cursor launched Origin code hosting
Cursor entered code hosting with Origin. It includes repositories, pull requests, code browsing, agents, and a staged early-beta rollout across paid plans.
An Origin-hosted repository uses Origin as its source of truth. You can also sync an existing GitHub repository while GitHub remains the source of truth.
SpaceX closed the Cursor acquisition
SpaceX closed the Cursor deal on August 14, 2026. Cursor sits in the same company as Grok Bot. That is why a Grok Bot login is a Cursor login, and why a Bot can hand work to a Cursor cloud agent.
xAI had already taken the SpaceXAI name in July. The August close is the one that put the coding IDE and the always-on Bot under one roof.
Z.ai released GLM-5.3
GLM-5.3 is Z.ai's coding model for long-running engineering work. It can keep a very large amount of code in view at once.
It was available first through the GLM Coding Plan. Two weeks later Z.ai released the model so anyone can download and run it.
Zed announced Delta
Delta is a separate coding environment from Zed. DeltaDB keeps agent conversations and worktrees together, so you can follow the work while it happens and review the reasoning beside the code.
The August announcement opened a private beta. Delta entered public beta on September 16. It works with existing Git repositories, and normal commits and pushes still work.
xAI launched Grok Bot
Grok Bot entered public beta on August 11, 2026 for SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium. A Bot is a named worker with its own memory, skills, and routines. It runs on a persistent cloud computer that keeps going after you close the laptop.
All of your Bots share that one machine. They share files, browser sessions, and credentials. I use it for jobs that start behind a login, not for work that should end in a pull request.
Meta released Muse Glimmer
Muse Glimmer is an open model from Meta for agents that run on your own computer. It can read text and images, use tools, and recover when a tool fails.
You can download it for free, including smaller versions that fit on a well-equipped laptop.
Cloudflare introduced Kitesurf
Kitesurf is a web browser built for AI agents, running on Cloudflare instead of on Chrome. Existing browser automation tools can talk to it without changes.
It is in beta. It uses far fewer resources than a normal browser, but it is slower and does not work with every website yet.
bb launched as a programmable IDE for coding agents
bb is an open-source, local-first IDE and orchestrator for coding agents. It launched with support for Codex, Claude Code, Cursor, and ACP agents while using the subscriptions you already have.
The interface is programmable too. You can ask an agent to create a plugin or add a feature to bb itself.
Cloudflare introduced @cloudflare/computer
Cloudflare Computer is an open-source project that gives an AI agent its own files and a place to run code, and keeps them around between sessions.
It is an early preview, with a lightweight option and a full Linux option.
Cloudflare announcement →A deep dive into Cloudflare Computer →
Anthropic released Claude Opus 5
Claude Opus 5 launched at the same price as Opus 4.8. It can work with very large projects, and it thinks before answering by default.
It became the default model on Claude Max and Anthropic's strongest model on Claude Pro.
Anthropic announcement →What I learned reading Claude's system prompts →
Moonshot launched Kimi K3
Kimi K3 is Moonshot's new large model. It understands text and images and can work with very long inputs.
Moonshot launched it as a hosted service first, then released the model for download later in July.
LM Studio launched Bionic
Bionic is a separate app from LM Studio for coding and document work. It can use a local model, a model on another machine through LM Link, or LM Studio Secure Cloud.
Code projects can inspect repositories, edit files, search code, and show inline diffs. Work projects add a sandbox, checkpoints, previews, and web search.
GPT-5.6 Sol became generally available
OpenAI previewed the GPT-5.6 family on June 26, 2026: Sol as the flagship, Terra for everyday work, Luna as the cheap fast one. A small partner group got API and Codex access first.
General availability landed on July 9, 2026 in ChatGPT, Codex, and the API. Sol is the model I still pick when the task is specified well enough that I want the code written carefully.
xAI became SpaceXAI
On July 6, 2026 the xAI account on X switched to @SpaceXAI and shipped a new logo. SpaceX had already bought xAI in February. July was the public name change, not the ownership change.
Grok, X, and later Cursor all sit under that brand. Product pages still say xAI in a lot of places, including the Grok Bot docs.
Vercel launched eve
eve is Vercel's open-source framework for building and running AI agents. An agent is a directory with Markdown instructions and TypeScript tools, plus optional skills, subagents, channels, and schedules.
Durable sessions, sandboxed execution, approvals, tracing, and evals come with the framework. The first release was a public preview.
Vercel announcement →Vercel eve: an open framework for building AI agents →
Anthropic released Claude Fable 5
Fable 5 is a new tier above Opus. Anthropic called it a Mythos-class model made safe enough for general use. It costs about twice as much as Opus.
Mythos 5 is the same model with some safeguards off, available only through Project Glasswing. On the public product, some Fable 5 requests can still get handed to Opus 5.
This is the model that made the named-tier thing real for me. Opus stayed in the mix. Fable became the one I use when the job is judgment, not typing.
Anthropic released Claude Opus 4.8
Claude Opus 4.8 kept the same price as Opus 4.7, plus a fast mode that answers much quicker for a higher price.
Claude Code also got a preview of dynamic workflows, where one session can coordinate many helper agents working in parallel on a bigger job.
Anthropic announcement →What I learned reading Claude's system prompts →
OpenAI released GPT-5.5
GPT-5.5 rolled out in ChatGPT and Codex, with a faster mode in Codex.
Developers got it through the API one day later.
Anthropic released Claude Opus 4.7
Claude Opus 4.7 became available in all Claude products and the API, at the same price as Opus 4.6.
It added a higher effort setting for harder problems and got better at reading large images.
Anthropic announcement →What I learned reading Claude's system prompts →
OpenAI expanded Codex beyond coding
Codex gained background computer use, image generation, and more than 90 plugins. OpenAI also added GitHub review comments, multiple terminals, an in-app browser, and SSH development boxes in alpha.
Automations could resume an existing thread and keep working over days or weeks. Memory launched as a preview.
Cursor released self-hosted cloud agents
Cursor cloud agents can now run on your own machines instead of Cursor servers. Your machine connects to Cursor on its own, so there is nothing to open up in your network.
Cursor still handles the models and the planning, while the commands run on your machine. There is also a setup for companies that want to run many of them.
Cursor announcement →Run Cursor cloud agents on your own Mac mini →
Anthropic released Claude Opus 4.6
Claude Opus 4.6 was the first Opus model that could work with very large projects in one go, as a beta feature.
The same release introduced Agent Teams in Claude Code as a research preview. One lead agent could split a job among several teammates and combine their work.
Anthropic announcement →What I learned reading Claude's system prompts →
OpenAI released GPT-5.3-Codex
GPT-5.3-Codex was built for long-running coding and research tasks. OpenAI said it ran 25% faster than GPT-5.2-Codex and could keep working across larger projects.
It launched across the Codex app, CLI, IDE extension, and cloud for paid ChatGPT plans.
OpenAI launched the Codex app
The first Codex desktop app launched on macOS. It put several agents in one workspace, let them work in parallel, and included skills plus scheduled Automations.
This was the standalone Codex app. OpenAI later moved Codex into the ChatGPT desktop app.
SpaceX acquired xAI
SpaceX completed its acquisition of xAI on February 2, bringing xAI, X, Starlink, and SpaceX into the same company.
SpaceX described orbital data centers as a long-term goal. The company later adopted the SpaceXAI name and bought Cursor.
Moonshot released Kimi K2.5
Kimi K2.5 is a Moonshot model for coding and tool use that understands both text and images.
It also introduced Agent Swarm, where one model coordinates several agents working on parts of the same job.
Anthropic released Claude Opus 4.5
Claude Opus 4.5 launched in Claude, the API, and the major cloud platforms, at a much lower price than earlier Opus models.
Claude Code gained an editable planning workflow with the same release. You could change the plan before the agent started writing code.
Google launched Gemini 3 and Antigravity
Google released Gemini 3 Pro through the Gemini app, Search, AI Studio, and Vertex AI. The same announcement introduced Antigravity, a coding environment built around agents.
Antigravity entered a free public preview. Its agents could work across the editor, terminal, and browser.
Google announcement →Choosing between an IDE, CLI, and vibe coding tool →
OpenAI released GPT-5.1
GPT-5.1 launched in ChatGPT as Instant and Thinking. Both models adjusted how much reasoning they used to match the task.
The ChatGPT rollout started on November 12. API access followed one day later.
Anthropic released Claude Sonnet 4.5 and the Agent SDK
Claude Sonnet 4.5 became Anthropic's new mid-tier model for coding and agent work.
Anthropic also released the Claude Agent SDK, checkpoints in Claude Code, and a VS Code extension.
Anthropic announcement →What I learned reading Claude's system prompts →
Anthropic released Claude Opus 4.1
Claude Opus 4.1 shipped as an upgrade for coding and agent work. It became available in Claude, Claude Code, the API, Amazon Bedrock, and Google Vertex AI.
Pricing stayed the same as Opus 4. Anthropic later replaced it with newer Opus releases.
Anthropic announcement →What I learned reading Claude's system prompts →
Vercel released AI SDK 5
AI SDK 5 added typed chat support for React, Svelte, Vue, and Angular. It also expanded tool handling, agent-loop controls, and speech support.
This was the main TypeScript framework release from Vercel that summer. The current AI SDK has moved several major versions beyond it.
Z.ai released GLM-4.5
GLM-4.5 is a free, downloadable model from Z.ai for coding and agent work. It can answer quickly or take time to think first.
Z.ai offered it both as a hosted service and as a download. Later GLM releases replaced it in ZCode.
OpenAI launched ChatGPT agent
ChatGPT agent combined deep research, Operator, a visual browser, a terminal, and connected apps. It could research a task, use websites, run code, and take actions from the same conversation.
This release also absorbed the standalone Operator workflow. Deep research remained available as a separate feature.
OpenAI announcement →How AI agents should handle passwords →
Moonshot released Kimi K2
Kimi K2 is a large model from Moonshot for coding and tool use, released for free download.
An updated version followed in September. Kimi K2 Thinking and Kimi K2.5 came later.
xAI released Grok 4
Grok 4 launched with built-in tool use and live search, in the Grok app and the API. xAI also introduced Grok 4 Heavy, a pricier version where several agents work on the same problem.
Grok 4 Fast followed in September, then Grok 4.1 in November.
Anthropic made Claude Code generally available
Claude Code reached general availability three months after its research preview. Anthropic added IDE integrations, background tasks, and broader access for developers.
The release arrived with Claude Opus 4 and Sonnet 4, which became the new models behind Claude Code.
Anthropic announcement →Choosing between an IDE, CLI, and vibe coding tool →
OpenAI released o3 and o4-mini
OpenAI released o3 and o4-mini as reasoning models that could use ChatGPT tools while they worked. They could browse the web, run Python, inspect uploaded files, and reason over images.
Both models later changed availability, so this entry describes their launch state.
OpenAI launched Codex CLI
The first Codex CLI was an open-source terminal agent installed through npm. It could inspect a repository, edit files, and run commands from the terminal.
OpenAI later replaced most of that original implementation with a Rust codebase. The product kept the Codex CLI name.
OpenAI released GPT-4.1
OpenAI launched GPT-4.1 in three sizes for developers. All of them could read very large inputs, like whole codebases.
The family focused on coding and following instructions carefully. It was only available through the API at launch.
Anthropic released Claude Code in research preview
Anthropic introduced Claude Code as a terminal coding agent alongside Claude 3.7 Sonnet. It could inspect a codebase, edit files, run commands, and work with Git from the terminal.
The first release was a limited research preview. Claude Code reached general availability three months later, on May 22.
Anthropic announcement →Choosing between an IDE, CLI, and vibe coding tool →
xAI announced Grok 3 Beta
Grok 3 Beta included Grok 3, Grok 3 mini, and Think variants that spent more time reasoning. xAI also introduced DeepSearch for researching information across the web.
The models were still in beta, and API access was only promised at launch.
GitHub previewed Copilot agent mode
Copilot agent mode could edit several files, inspect its own errors, run another iteration, and suggest terminal commands. The first preview required VS Code Insiders.
This was the point where Copilot moved beyond inline suggestions and chat toward a coding-agent loop inside the editor.
OpenAI released o3-mini
o3-mini was a small, fast reasoning model. You could choose how long it thought before answering.
It focused on text and code, and it could not read images.
DeepSeek released DeepSeek-R1
DeepSeek released R1 as an open-weight reasoning model with website and API access. The weights and code used the MIT license.
The release also included six smaller distilled models. DeepSeek compared R1 with leading closed reasoning models using its own evaluations.