Key takeaways
- Act: If your automations call Claude Opus 4.8 with fast mode and you need to keep fast mode, test Claude Opus 5.5 and switch before 30 November. Otherwise, check whether Opus 4.8 at standard speed is fast enough. If you administer a Claude Team or Enterprise organization with Skills turned off, decide before 30 November whether plugins in Claude Desktop should keep working.
- Plan: If you want to run a large model on your own infrastructure or with a European provider, add Mistral Large 4 to your shortlist and read the licence condition first.
- Plan: If you use the Claude reporting endpoints that are now legacy, plan the move to the new Usage and Cost Analytics API. Before you use message threads on the Claude API, check whether you rely on zero data retention.
- Watch: Gemini in Slides, Sheets and the Gemini app. Check which features your Workspace edition includes.
- Watch: Gemini 4 Carbon. Google gives no date for broader access.
01 / 06
Claude: two deadlines on 30 November
What happened. On 30 October 2026, Anthropic deprecated fast mode for Claude Opus 4.8, with removal on 30 November 2026. After removal, requests to claude-opus-4-8 with speed: "fast" return an error. To continue using fast mode, Anthropic says to migrate to Claude Opus 5.5. (Claude Platform release notes) This removes only the speed option: Opus 4.8 itself stays available at standard speed. According to Anthropic's price table, fast mode for Claude Opus 5.5 costs 8 US dollars for input and 40 US dollars for output per million tokens. For Claude Opus 4.8, the table lists 10 and 50 US dollars. (Anthropic: Fast mode)
In the Claude apps, Anthropic is bringing plugins, skills and connectors together on one Plugins page. For Free, Pro, Max, Ultra and Team plans, the change has been rolling out gradually since 29 October and becomes generally available on 9 November. For Enterprise plans, it is on by default from 30 November. Existing connectors, skills and plugins stay installed. From 30 November, in Team and Enterprise organizations where Skills are turned off, plugins are also blocked in the Claude Desktop app. Admins who want to keep plugins must turn on Skills before 30 November. Owners do this under Organization settings > Library > Policy, where "Cloud code execution and file creation" must also be on. (Claude Help Center: Manage plugins, skills, and connectors from one Plugins page)
What it means for SMEs. If a service provider built automations for you on the Claude API, a forgotten fast-mode setting can stop a workflow on 1 December. In the Claude apps, a decision that you made earlier about Skills now also decides about plugins.
Our recommendation: act. Ask your team or service provider whether any workflow calls claude-opus-4-8 with speed: "fast". If so, decide whether the workflow needs fast mode. If it does, test it with Claude Opus 5.5 on real cases and switch before 30 November. If it does not, remove the fast-mode setting and check the response times at standard speed. In a Team or Enterprise organization, check whether Skills are on, and decide whether plugins should keep working. Record which plugins are allowed in your AI usage policy.
02 / 06
Mistral Large 4: open weights with one licence condition
What happened. On 29 October 2026, Mistral published the weights of Mistral Large 4, according to its news page. We reported on the public preview in week 41. According to the model card on Hugging Face, it is a multimodal mixture-of-experts model with about 1 trillion parameters, of which 49 billion are active per token (52 billion including embedding and output layers), and a context window of 1 million tokens. (Hugging Face: Mistral-Large-4-1T-A52B) The licence is the Mistral Ecosystem Open-Weight License, a modified MIT licence. According to Mistral, the model is free to use, modify, fine-tune and commercialize, with one condition: businesses that operate a model-as-a-service offering must obtain a separate commercial licence from Mistral before using the model commercially. In Mistral's API, the model costs 0.70 US dollars for input, 0.07 US dollars for cached input and 2.10 US dollars for output per million tokens. (Mistral: Le Chonk, built for coders, trusted by defenders)
What it means for SMEs. Companies that want a large model outside the US providers have one more option. Self-hosting a model of this size requires substantial computing resources. Compare the hardware and operating costs with a hosted offer. If you offer the model to your own customers as a service, check the licence condition first.
Our recommendation: plan. If data location or provider independence matters to you, test Mistral Large 4 with twenty real cases against your current model, as described in Your own AI benchmark. Before you decide between local operation and a European cloud, read Local AI or EU cloud.
03 / 06
Claude API: new cost report and message threads without zero data retention
What happened. On 29 October 2026, Anthropic launched the Usage and Cost Analytics API in beta. It reports token usage and cost for Claude Console and Claude Enterprise organizations. You can filter and group the results by model, workspace, API key and the user or service account behind a request. The existing Usage and Cost API and four report endpoints of the Claude Enterprise Analytics API are now legacy, but Anthropic says that all of them remain supported. (Claude Platform release notes)
On 2 November 2026, Anthropic added message threads to the Messages API, in beta. The API keeps the conversation on the server, so each later request sends only the new messages. Anthropic says this gives smaller requests and a lower time to first token. A thread expires 5 minutes after its last request and is deleted no more than 1 hour after its last use. (Claude Platform release notes) According to the documentation, the thread content is held encrypted in memory and is not written to durable storage. The feature is not eligible for zero data retention (ZDR). (Anthropic: Message threads)
What it means for SMEs. With the new cost report, you can assign AI costs to projects, departments or clients, for example by API key. Existing reports keep working. Message threads can make chatbots and agents with long conversations respond faster. But Anthropic keeps the thread on its servers and deletes it no more than 1 hour after its last use, also for companies that have an agreement for zero data retention.
Our recommendation: plan. If you use the legacy reporting endpoints, assess the switch to the new cost report in your next planning cycle. Give each project its own API key so that the costs can be separated. Before you use message threads, check with your data protection officer whether you rely on zero data retention. What to check for chatbots is described in Chatbot data protection.
Got a workflow like this?
Tell me about it. Honest take and a rough price, usually within a day.
04 / 06
Google Workspace: Gemini edits whole presentations
What happened. On 30 October 2026, Google announced that Gemini in Google Slides can edit several slides or an entire presentation from the side panel. The feature is available in English only at launch. The rollout started on 29 October. Google lists Business Standard and Plus, Enterprise Standard and Plus, Google AI Pro and Ultra, and Google AI Pro for Education. (Google: Edit entire presentations with Gemini in Google Slides) On 28 October, Google announced that "Fill with Gemini" in Google Sheets shows a citation for each cell that it fills from web sources, and that a request can now cover up to 3,000 cells. The citations are available in Google Sheets on the web. The editions are the same, plus AI Expanded Access. (Google: Fill with Gemini now provides cell-level citations) On 29 October, Google announced that users can create, preview and edit files directly in a Gemini chat. PDFs work on the web and on mobile, and all other formats, including presentations and spreadsheets, on the web only. Google lists all Workspace customers, Workspace Individual subscribers and personal accounts, in the languages in which the Gemini app is available. (Google: Create, preview, and edit files directly in the Gemini app)
What it means for SMEs. Teams that build sales or project presentations in Slides can let Gemini change a whole deck in one step, at first only in English. With citations per cell, you can check where a value in a spreadsheet came from before you use it.
Our recommendation: watch. Check which of these features your Workspace edition includes. Before you use values that Gemini filled from the web, open the citation and check the source.
05 / 06
Gemini 4 Carbon: announced, first for selected cyber defenders
What happened. On 28 October 2026, Google announced Gemini 4 Carbon, its new frontier model. Starting that week, Google makes the model available to a subset of cyber defenders and gradually gives more trusted partners access through its Fairwind Program. Developers, businesses and consumers are to follow "as soon as possible", without a date. For the launch, Google states an introductory price of 2 US dollars for input and 10 US dollars for output per million tokens, with cached input at 95% off the input price. After the introductory period, 4 and 20 US dollars apply. (Google: Gemini 4 Carbon)
What it means for SMEs. First access is limited to selected cyber defenders and additional trusted partners, and Google gives no date for broader access. Announced prices and benchmark results do not tell you how well a model handles your own tasks.
Our recommendation: watch. When Gemini 4 Carbon becomes available, test it with your own cases, as with any new model. How the major assistants differ for business use is compared in ChatGPT, Claude or Gemini.
06 / 06
In brief: the other providers
| Provider | What is confirmed | Relevance for SMEs |
|---|---|---|
| Anthropic | The maximum request size for the Messages API and the Token Counting API rose from 32 MB to 128 MB on the first-party Claude API on 28 October (release notes) | Watch: for large documents and many images in one request |
| OpenAI | In ChatGPT Enterprise and Edu, automatic switching from Instant to higher thinking levels and its workspace setting were retired on 28 October. Users select the available thinking levels themselves. Workspaces configured for specialized medical routing are excluded, and legacy models can still route for safety or personal-finance features (Enterprise and Edu release notes) | Watch: tell your users when a higher thinking level is worth it |
| OpenAI | Eligible ChatGPT Business workspaces can activate promotional API credits in a linked API organization since 28 October. These credits are separate from ChatGPT usage (Business release notes, terms) | Plan: check in your account whether you are eligible |
| OpenAI | Pro 100 and Pro 500 receive twice their previous Work and Codex allowance since 29 October. OpenAI canceled the planned reduction for Pro 200 (ChatGPT release notes) | Watch |
| Microsoft | Reminder: from 1 December 2026, new Microsoft 365 Copilot and Copilot Business offers can be bought without a Microsoft 365 base subscription. For eligible usage-based experiences, new Copilot Business subscriptions through partners (CSP) default to pay-as-you-go billing, with an adjustable default limit of 4,000 Copilot Credits per user per month. Germany is initially excluded from this billing default (Partner Center announcements, October 2026). Purchasing details in week 43, billing details in week 41 | Plan: if you order new Copilot licences in December |
We also checked Meta, Aleph Alpha, DeepSeek, Qwen, Cohere, AWS Bedrock, n8n, Make and Zapier as well as the European Commission and the AI Office, the RTR KI-Servicestelle, the Bundesnetzagentur and the data protection authorities in Austria and Germany, and selected no item from them for this issue. Left out on purpose: benchmark claims, stock market listings and funding news.
If you want to know what one of these changes means for your existing workflows, we are happy to talk it through in an AI consulting call.
Other issues of the AI briefing: week 41/2026 · week 42/2026 · week 43/2026 · week 44/2026.
