GPT-6 gets smarter charts, Haiku 5.5 gets 75% cheaper, Mistral goes trillion-parameter, and India starts a 30-day countdown on AI regulation.
OpenAI's GPT-6 now builds charts and interactive forms inside your chat
OpenAI added an "Intelligent UI" mode to GPT-6 in ChatGPT. When you ask for data or a comparison, it now returns interactive charts, forms, and calculators — not just a text answer. The feature is live immediately for Pro, Plus, Business, and Enterprise users, with Free and Go tier access from October 8.
Why it matters: The shift from text output to interactive output makes AI feel less like a search engine and more like a working tool. If you use ChatGPT for planning, analysis, or decision-making, expect more of that work to happen inside a single chat window instead of switching between apps.
Claude Haiku 5.5 is 75% cheaper — and Anthropic says it's faster too
Anthropic released Claude Haiku 5.5 this week at $0.10 per million input tokens and $0.50 per million output tokens — roughly 75% cheaper than the previous Haiku 4.5. The context window stays at 100,000 tokens.
Why it matters: Cheaper models mean smaller teams and solo developers can build more with the same budget. If you run Claude for business automation, content generation, or student projects, your costs just dropped significantly.
Mistral's new model has a trillion parameters — but only runs 49 billion at a time
French AI company Mistral launched Large 4 preview: a natively multimodal mixture-of-experts model with 1 trillion total parameters, of which only 49 billion are active at any given moment. That design keeps it fast and efficient despite the scale. It can process text, images, and code in the same conversation.
Why it matters: Mixture-of-experts is the architecture behind most of the world's fastest frontier models. Mistral competing at this scale keeps the top end of the market from becoming a two-company race between OpenAI and Anthropic — and that competition usually benefits users through lower prices and faster improvements.
India's government says a binding AI regulation paper is coming within 30 days
IT Minister Ashwini Vaishnaw announced on October 8 that MeitY will release a consultation paper on AI regulation within a month. Priorities named: deepfakes, AI safety, cybersecurity, and responsible technology. One report also says India is simultaneously procuring 10,000 more GPUs for public-interest AI workloads.
Why it matters: India has over 1.4 billion people and one of the world's fastest-growing AI sectors. Once this consultation paper drops, businesses and developers building AI products in India will have the clearest picture yet of what the legal guardrails will look like.
AI companies took eight of the ten biggest US startup deals last week — $2.86B in seven days
For the week of September 26 to October 2, eight of the ten largest US startup funding rounds went to AI companies, totalling $2.86 billion. Instinct AI raised $1 billion in a Series C. Crusoe Energy — which builds AI data centre infrastructure — raised $350 million in a Series F. Across all of 2026, AI startups have reportedly raised $74.1 billion across 77 rounds.
Why it matters: Investor appetite for AI is not cooling. The money flowing in right now is going to infrastructure and vertical software — not experimental ideas — which tells you where the durable market is forming and where the jobs will be.
NVIDIA open-sourced a robot AI model that learns the physical world, not just text
NVIDIA released Long-WAM, a world-action model designed for robots. Unlike language models trained on text, Long-WAM learns how the physical world behaves — so a robot can use it to plan and carry out real-world actions. Code and model weights are released under Apache 2.0, meaning anyone can build on it without a licence fee.
Why it matters: Language models made software smarter. World-action models could do the same for physical automation — from warehouse robots to assistive devices. Open-source access means smaller teams can experiment without needing a full hardware lab.
India's AI regulation: 30 days to a blueprint that could shape 1.4 billion people
India's consultation paper on AI safety and deepfakes is due within 30 days. The country has already provisioned 38,000+ GPUs, embedded AI literacy in schools from Class 3, and attracted a $500B economic forecast for AI's impact by 2030. The regulation paper will be the first time legal guardrails formally follow that infrastructure investment. Whatever India decides here is very likely to be referenced by governments across South and Southeast Asia.
Africa's first open multilingual model + Mistral's $3.5B raise
Vambo AI in South Africa released MORENA, an open-weight model trained from scratch on 12 African languages including Shona, Zulu, and Hausa — the first model of this kind at this scale for languages that most global AI systems barely cover. Separately, Mistral AI closed a $3.5 billion Series D, one of the largest-ever European AI funding rounds, reinforcing its bid to stay competitive at the frontier.