Key takeaways
- Anthropic launched Claude Fable 5.2 and Mythos 5.2 on November 9, reported to be cheaper than OpenAI's GPT-6 Astra.
- Gemini 4 Carbon is open to paying developers at $2 / $10 per million tokens until November 30.
- OpenAI's Control Room can cut off a stopped AI agent job in about 35 seconds.
- Claude Meetings (beta) lets Claude join Google Meet calls; Anthropic also started its IPO roadshow.
- India's IndiaAI Mission launched EKA-1 (22 Indian languages) and offline Dhīr models.
AI moved fast again this week. Two new top models, a cheaper Google model for developers, a "kill switch" for AI agents, and Claude joining your video calls. If you only have five minutes, this is the AI news this week that matters, in plain English, with a link to the original source for each item.
For each story you'll find what happened, why it matters to you (shop owner, freelancer, student or developer) and the source.
1. Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2#
What happened: On November 9, Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2, which it calls its most capable models so far for hard technical work in coding, finance and science. Fortune reports they launched at a lower price than OpenAI's latest frontier model, GPT-6 Astra.
Why it matters: better models at lower prices mean AI coding help, data analysis and long reports get cheaper for every business building on Claude. If your app or chatbot runs on an older Claude model, test the new ones on your own tasks before you switch.
Source: Anthropic announcements · Fortune
2. Google opens Gemini 4 Carbon to paying developers#
What happened: On November 10, Google made Gemini 4 Carbon available on the paid Gemini API (Google AI Studio, model ID gemini-4-carbon). Launch pricing is $2 per million input tokens and $10 per million output tokens until November 30, after which it is reported to rise to $4 and $20.
Why it matters: if you're a developer, the launch price is a cheap window to test it against the model you use now. Businesses using AI chatbots or document tools built on Gemini may see better answers. Budget for the December price change before you move a live product.
Source: Technology.org · Google for Developers
3. OpenAI builds "Control Room", a 35-second kill switch for AI agents#
What happened: OpenAI described Control Room, a central system that can stop model output for its internal research jobs, including AI agents. Every job needs permission from Control Room before the model answers. When a job is stopped, it gets no more output within about 35 seconds. If Control Room itself fails, all requests are blocked ("fail closed").
Why it matters: AI agents now run code, browse and send messages on their own. A reliable stop button is becoming a basic safety need. If you use AI agents in your business, ask your vendor one question: how fast can we stop it?
Source: Interesting Engineering · BankInfoSecurity
4. Claude can now join your meetings (beta)#
What happened: At a November 10 event, Anthropic showed new work tools. Claude Meetings lets Claude join a video call as a visible participant, in beta for Team and Enterprise plans, starting with Google Meet. Fortune also reports Claude Whiteboards and Claude Sheets (beta, Sheets exports to Excel/CSV), Claude Data for paid plans, and more than 9,000 apps available as Claude Connectors.
Why it matters: meeting notes, action items and spreadsheets are where most office time goes. These tools aim straight at that work. Teams already on Claude can try them now; others can expect similar features in their AI assistants soon.
Source: Fortune
5. Anthropic starts its IPO roadshow#
What happened: On November 10, Anthropic announced it launched the roadshow for its proposed IPO (initial public offering) of Class A stock. Fortune reports a target range of $70–$78 a share, which could value the company at roughly $1.4 trillion.
Why it matters: it's one of the biggest tech listings ever and shows how much money is flowing into AI. For everyday users nothing changes today, but public companies face more pressure on pricing and product speed. This is news, not investment advice.
Source: Anthropic announcements · Fortune
6. India: IndiaAI Mission launches EKA-1 and Dhīr models#
What happened: Last week (November 4), the IndiaAI Mission launched three homegrown models. EKA-1 by Soket Labs has 26.3 billion parameters and covers all 22 scheduled Indian languages plus English. Dhīr 2B and Dhīr 5B by ZenteiQ can run offline, even in air-gapped systems, for document-based work.
Why it matters: AI that understands Indian languages and runs without the internet is useful for government offices, banks, schools and small businesses that can't send data to foreign clouds. Public access details are not out yet.
Source: News On AIR
The big picture this week#
- Prices are falling, power is rising. New top models are launching cheaper than the ones they replace.
- AI is moving into office work. Meetings, sheets and whiteboards are the new battleground, not just chat.
- Safety is getting practical. Kill switches for agents are now a product feature, not a research idea.
Two prompts to use this news#
Use these with ChatGPT, Claude or Gemini to make sense of any AI news, or to pick the right model for your budget.
1. Explain any AI news simply
Explain this tech news to me in simple English, like I'm a busy small business owner.
News: {{PASTE_HEADLINE_OR_ARTICLE}}
1. What happened, in 2 sentences.
2. Why it matters for someone like me ({{MY_ROLE}}).
3. One thing I should do this week because of it, or "nothing" if it doesn't affect me.
4. What is still unconfirmed or hype.
Keep it under 150 words. No jargon.
2. Choose an AI model API by cost
Help me choose an AI model API for my project based on cost.
My use case: {{USE_CASE}}
Expected volume: {{REQUESTS_PER_DAY}} requests a day, about {{INPUT_TOKENS}} input and {{OUTPUT_TOKENS}} output tokens each.
Models I'm considering and their prices per million tokens: {{MODEL_PRICES}}
1. Calculate my monthly cost for each model in a table (show the math).
2. Say which one gives the best value for this use case and why.
3. List 3 tests I should run on my own data before I switch.
Don't invent prices. Use only the prices I gave you.
Want more? See our guide to the best AI tools for small business in India and how agentic AI is changing software development.
Put this week's AI to work in your business#
Larwell builds AI chatbots, WhatsApp automation, billing software and websites for Indian businesses. Want a chatbot on Gemini or Claude, or an AI agent with a proper stop button? Talk to us or see our services. A fresh AI news roundup comes out every week on the Larwell blog.
