Upgrades

Anthropic launched Salesforce in Claude in beta, a plugin bringing a seller's accounts, opportunities and pipeline into Claude with 37 pre-built sales skills for call prep, deal review, pipeline dashboards and forecasts.

Who can use it Beta on all paid Claude plans, but only for organizations that Salesforce approves through its own beta sign-up.

The Messages API can now compact a conversation on request: send the compaction parameter and the API returns a signed compaction block summarising the messages, which later requests send in place of those messages, optionally keeping recent turns word for word.

Who can use it Beta on the Claude API with the compact-2026-09-04 beta header.

The Compliance API's local session endpoints now also return transcripts of Claude in Chrome sessions, under the product_surface value claude_in_chrome.

Who can use it Beta for Claude Enterprise organizations, using the existing Compliance Access Key with the read:compliance_user_data scope.

The @cloudflare/voice package now reports a typed summary for every speech or text turn, including a terminal outcome (completed, no output, content filtered, model error and others) and per-stage timings from speech-to-text through to audio playback.

Who can use it Available now in @cloudflare/voice v0.4.0 with a compatible Agents SDK version; Cloudflare names no plan restriction.

Cloudflare AI Search can now index R2 objects that have no filename extension, as long as they carry a supported Content-Type, while still enforcing file-type validation.

Who can use it Available now for AI Search data sources backed by R2; Cloudflare names no plan restriction.

A new rejectIfBusy option makes synchronous Workers AI inference requests fail when capacity is unavailable, instead of waiting in a capacity queue.

Who can use it Available now as an option on the Workers AI binding and in the REST API request body; Cloudflare names no plan restriction.

GitHub Copilot's auto model selection now offers three tiers, efficiency, balance and intelligence, that change how auto weighs cost, quality and response time before picking a model for each prompt.

Who can use it Rolling out now in Visual Studio Code, Copilot CLI and the GitHub Copilot app. Usage is billed by whichever model auto selects, regardless of tier; paid subscribers keep a 10% discount on auto usage.

GitHub's AI Scan for pull requests can now find security vulnerabilities on repositories without CodeQL default setup enabled, so it runs more broadly across an organization's eligible repositories.

Who can use it Public preview for organization-owned and personal repositories on github.com for GitHub Advanced Security customers; GitHub Enterprise Server is not supported.

The Copilot impact dashboard now shows how many active users engaged with each Copilot feature on at least two days in a 28-day period, from code completion and agent edits to code review, cloud agent, CLI and the Copilot app.

Who can use it Dashboard for enterprise administrators; the same breakdown is added to enterprise and organization 28-day report APIs.

Google introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, voice models built for near-real-time reasoning, background tool calls, and live multi-step task narration.

Who can use it 3.8 Live is rolling out today for developers in the Gemini API and Google AI Studio, and for everyone in Search Live. 3.8 Live Extended Thinking is in private preview for Gemini Enterprise.

Mistral and Mozilla announced a partnership powering Firefox's Smart Window AI browsing assistant, beta, with Mistral models fine-tuned on regional languages and dialects.

Who can use it Live now for users in France and North America; Mistral says the UK and Germany will follow later this year.

Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are now available through Vercel AI Gateway for real-time audio conversation apps, including background tool calls and multi-step reasoning that runs in parallel with speech.

Who can use it Available now through AI Gateway using the model IDs google/gemini-3.8-live and google/gemini-3.8-live-extended-thinking; Vercel names no plan restriction.

Mem0 is now a native Vercel Marketplace integration, giving AI agents and apps long-term memory that persists user preferences, facts and context across sessions, with a scoped API key provisioned automatically.

Who can use it Installable from the Vercel Marketplace with billing on the Vercel invoice. Mem0 offers a free plan with usage limits and a $20-a-month plan for higher volume.

Every Vercel Sandbox now includes 64 GB of storage, up from 32 GB, covering sandboxes built from managed or custom images and giving more room for large repositories and storage-intensive agent tasks.

Who can use it Applies automatically to new sandboxes on the current SDK and CLI versions; Vercel names no plan restriction.

OpenAI's GPT-Live 1, a full-duplex voice model that can listen and speak at the same time, is now available through Vercel AI Gateway, with client delegation to hand deeper work to any text model on the gateway.

Who can use it Available now through AI Gateway; requires AI SDK 7 and @ai-sdk/openai 4.0.67 or later.

Harbor, the open-source harness behind Terminal-Bench, can now run each eval trial in its own isolated Vercel Sandbox microVM, with network policy and optional credential injection enforced at the sandbox firewall.

Who can use it Available now by passing --env vercel to harbor run; Vercel names no plan restriction.

skills 1.7.0 adds Notion skills databases as an install source, so teams can write and review agent skills as Notion pages and install them into any agent the CLI supports without a Git repository.

Who can use it Available in skills@1.7.0; authentication uses the Notion CLI, and the Notion workspace must allow personal access tokens.

GLM 5.3 FlashX, a high-speed serving option for Z.ai's multimodal coding model at about 200 tokens per second, is now available through Vercel AI Gateway.

Who can use it Available now through AI Gateway with the model ID zai/glm-5.3-flashx; Vercel names no plan restriction.

Custom AI Gateway costs can specify separate cache-read and cache-write token rates while avoiding double-counting providers that report cache tokens differently.

Who can use it Available through AI Gateway custom costs; Cloudflare names no plan restriction in the changelog.

R2 can log successful object reads, writes, lists, multipart uploads and deletes across its APIs, dashboard, Workers bindings and public buckets into Workers Observability.

Who can use it Generally available for non-jurisdictional R2 buckets. Delivery is asynchronous and best effort, so Cloudflare says not to treat it as a complete activity record.

Cloudflare removed the old compressed 3 MB Free and 10 MB Paid limits and now checks one 64 MiB uncompressed bundle limit.

Who can use it Free and paid Workers plans.

GitHub announced Astra as generally available across Copilot's editors, CLI, coding agent, app, website and mobile surfaces on 4 September, while saying rollout would be gradual.

Who can use it Copilot Pro+, Max, Business and Enterprise. Usage-based billing at provider list pricing; Business and Enterprise administrators can manage access through model policy.

Enterprise administrators can centrally set shell, file and network operations to deny, require approval or proceed, with team-specific policies that local settings cannot weaken.

Who can use it Generally available for Copilot Business and Enterprise in the Copilot app, Copilot CLI and Visual Studio Code sessions using Agent Host.

Copilot's app and command-line agent workflows now honour configured content exclusions, keeping excluded code out of their context.

Who can use it GitHub names the Copilot app and Copilot CLI, but the changelog does not name a plan restriction for this item.

Select up to 25 code-quality findings, assign them to Copilot, and receive a branch and pull request after the agent validates its changes.

Who can use it Repositories with GitHub Code Quality enabled on GitHub Team and GitHub Enterprise Cloud, including data-residency deployments. Uses AI credits.

OpenAI began its own broader Astra rollout on 10 September, following earlier limited and partner distribution, with new computer-use, coding, science and cybersecurity capabilities.

Who can use it A limited set of organisations today; Plus, Pro, Business and Enterprise, the OpenAI API, Azure and AWS Bedrock over the coming days. Free is not named. Astra Pro is limited to Pro, Business and Enterprise.

Sandboxes can run in all 20 Vercel compute regions, with project defaults and optional ordered failover regions.

Who can use it Region selection on every plan. Failover regions require Pro or Enterprise. Regional CPU and memory rates vary.

eve agents can load and update named memory slots across sessions, with storage providers and scopes that define who or what shares each slot.

Who can use it eve agents. The built-in file provider is available with `eve add memory/file`; on Vercel it uses a private Blob store. The changelog names no plan restriction.

Vercel Authentication can now require authorised Vercel sign-in across every deployment in a project, including production domains.

Who can use it Every Vercel plan at no additional cost. Per-project password protection is a separate $20 monthly option for Pro teams.

Ask Slackbot to make images, video, designs and PDFs with 70+ Adobe tools, using files already in the channel.

Who can use it Slack Business+ and Enterprise+ only.

New top-end Claude model for coding, knowledge work and long-running agents.

Who can use it API and the cloud resellers. On claude.ai: not on Free; Pro via usage credits, Max, Team and Enterprise included.

Re-reading context you have already sent costs $0.25 per million tokens. Anthropic estimates ~25% off typical bills and ~45% off heavy agent workloads.

Who can use it Anyone billed by token. Input and output prices unchanged.

Cursor's cloud agents run commands and edit files on machines you control, so code and secrets stay in your network.

Who can use it My Machines on individual accounts. Team Pools require Enterprise.

DeepSeek's first experimental model that reads images as well as text, published as downloadable weights.

Who can use it Open weights, ungated, MIT licence (~168GB). Paid API endpoint also available.

Copilot judges whether a change is ready and, if an admin allows it, formally approves it toward required-approvals.

Who can use it Public preview on Pro, Pro+, Max, Business, Enterprise. Approval is off by default.

New model for coding and multi-step agents, at 3.7 Flash's speed and price.

Who can use it Consumers on AI Pro and Ultra — not the free tier. Developers at $0.75/$3.75 per million, rising to $1.50/$7.50 on 1 Jan 2027.

The model picks which parts of a video to watch instead of sampling every frame, cutting token use by up to 88%.

Who can use it Developers now, in AI Studio and the Gemini API. Standard token pricing.

Talk to Gmail to find things, to Docs to draft by voice, to Keep to turn spoken notes into structured ones.

Who can use it Gmail and Keep Live on AI Plus, Pro and Ultra. Docs Live on Pro and Ultra only. Workspace business not yet.

Typing suggestions no longer appear in Word and Outlook unless you switch them on.

Who can use it Word web, Windows, Android, iOS and classic Outlook. Applies to free and paid alike.

Document OCR with paragraph-level bounding boxes, block labels and per-block confidence scores.

Who can use it Mistral API. A Premier commercial model — pay-per-use; self-deployment needs a commercial licence.

One switch turns off every AI feature in Firefox, present and future, with individual toggles beneath it.

Who can use it Free, all Firefox users.

Finds other PCs on your home network and spreads AI work across them instead of queuing on one graphics card.

Who can use it Free and open-source beta. Windows, macOS, Linux. Needs RTX 20-series or newer, an RTX PRO GPU, a DGX Spark, or an Apple M4 or newer. Works with Ollama and LM Studio.

Attach more than one Gmail, Calendar and Contacts account, so work and personal appear in one conversation.

Who can use it Plus, Pro, Business, Enterprise. Not Free.

ChatGPT can use tools a website publishes via WebMCP, without you connecting anything.

Who can use it ChatGPT Work and Codex, desktop app's built-in browser only — not the Chrome extension.

Mention open tabs as context or hand ChatGPT a browser task from those browsers.

Who can use it Desktop app users. Side chat works in Edge, Brave and Vivaldi but not Opera.

Splits a task between a cloud model and one running on your Mac, with an on-device gate that masks sensitive file content before anything leaves.

Who can use it Pro, Max and Enterprise. Apple silicon, macOS 15+, 24GB unified memory minimum.

Caps what each team member can spend on models per day, week or month, so one runaway agent cannot drain the budget.

Who can use it Needs CLI 59.6.2+. Keys made before this release are attributed to the team, not the person — rotate them or budgets will not apply.