Anthropic Rebuilt Its Entire Claude Lineup in 15 Days. Here’s What Actually Changed.
Between September 22 and October 7, Anthropic retired every model it sells and replaced each one. Opus, then Sonnet, then Haiku — three full lineup swaps in fifteen days. The headline number is the price: Opus 5.5 costs 40% less to run than Opus 5 while working at roughly the same level as Claude Fable 5.1.
That’s mostly not a smarter Claude — it’s a cheaper one, and cheaper is the story this week across the industry. Microsoft spent the same stretch splitting Copilot into three distinct products instead of one sprawling assistant. OpenAI added studying tools to the teen version of ChatGPT rather than new tricks to the adult one. None of these is a dramatic leap in raw capability — they’re moves to make existing capability cost less, run faster, or fit a narrower job better.
That matters for anyone paying per-token API bills, anyone whose team already lives inside Microsoft 365, or anyone with a teenager doing homework with a chatbot. Here’s what happened in each case, and what to actually do about it.
| Story | Lab | What it means | Date |
|---|---|---|---|
| Copilot splits into three products | Microsoft | Specialized tools, not one bot | Sep 25, 2026 |
| ChatGPT for Teens gets study tools | OpenAI | Flashcards, quizzes, planner | Oct 7–8, 2026 |
| Entire Claude lineup refreshed LEAD STORY | Anthropic | Cheaper and faster, across the board | Sep 22–Oct 7, 2026 |
Microsoft Split Copilot Into Three Products: Home, Code, and an “Always-On” Autopilot
One assistant becomes three specialized tools
Microsoft announced on September 25 that Copilot is no longer one tool — it’s three. Home merges Copilot Chat with the previously separate “Cowork” and builds Microsoft Office directly into that starting point. Code runs on the same underlying technology as GitHub Copilot, but is built for companies to use on their own tenant and data. Autopilot — internally known as “Scout” before this — is a persistent agent that keeps working on a task after you’ve closed the laptop; it moved into private preview at the end of September, while Home and Code are rolling out through Microsoft’s Frontier early-access program.
- Home merges Chat + Cowork, with Office built in
- Code runs on GitHub Copilot’s tech, scoped to your own company data
- Autopilot (formerly “Scout”) works unattended on ongoing tasks
- Microsoft is betting on specialized tools over one does-everything bot
- Home and Code are only reachable now via the Frontier early-access program — wide availability isn’t confirmed yet
Anthropic didn’t make Claude smarter this month. It made being smart cost 40% less — and did it three separate times in fifteen days.
ChatGPT for Teens Gets Flashcards and Quizzes. A College Planner Is Coming.
Study tools arrive before the adult tier gets anything new
OpenAI added two concrete study tools to ChatGPT’s teen tier on October 7–8: Flashcards, which build decks from uploaded class notes or a chosen topic, let students mark cards they’ve learned, and save decks to their Library; and Quizzes, which turn uploaded notes into interactive quizzes inside the chat itself. A College Planner is listed as “coming soon,” meant to help high schoolers in grades 10 through 12 track applications, deadlines, and financial aid. This builds directly on the teen tier OpenAI launched August 18, now the default for any account identified as belonging to someone under 18.
- Flashcards build decks from notes, live in the teen’s Library
- Quizzes turn uploaded notes into an interactive quiz in chat
- College Planner is announced but not live yet
- These tools are teens-only for now, not in the adult ChatGPT tier
Same Brain, Lower Bill: What Opus 5.5, Sonnet 5.5, and Haiku 5.5 Actually Change
Anthropic replaced its whole lineup in two and a half weeks
Anthropic didn’t ship one upgrade — it shipped three, back to back. Opus 5.5 (announced September 22) performs at the level of Claude Fable 5.1 on most work while costing 40% less than Opus 5 to run. Sonnet 5.5 (September 28) runs about 30% faster than Sonnet 5 and costs up to 30% less for most tasks. Haiku 5.5 (October 7) is, in Anthropic’s own words, the fastest, cheapest, and most capable small model it has built, aimed squarely at high-volume, cost-sensitive work. Line all three up and the pattern is clear: this was a cost-and-speed overhaul of the whole product line, not a single flashy upgrade.
- Opus 5.5: Claude Fable 5.1-level work, 40% cheaper than Opus 5
- Sonnet 5.5: ~30% faster and up to 30% cheaper than Sonnet 5
- Haiku 5.5: Anthropic’s fastest, cheapest, most capable small model yet
- All three tiers were refreshed within about two and a half weeks
- If you’re on API pricing, switching models can cut your bill with no code changes
Check Your Model Tier Before You Check Anything Else
If you or your team pay for Claude by the API token, check which model you’re calling. If you’re still pointed at Opus 5 or Sonnet 5, switching to the 5.5 versions should cut your bill (up to 40% on Opus-tier work, up to 30% on Sonnet-tier work) without changing your prompts or workflow. Test it on one workflow first before moving everything over.
If your business runs on Microsoft 365, ask your IT admin whether you’re enrolled in the Frontier early-access program — that’s the only way to get Copilot Home or Code right now. If you have recurring tasks you’d want handled without babysitting them, ask about Autopilot’s private preview specifically.
If you have a teenager using ChatGPT for schoolwork, point them at the Flashcards and Quizzes tools inside the teen tier — both work directly from uploaded class notes. If they’re a junior or senior, keep an eye out for College Planner; it isn’t live yet, but it’s built specifically for tracking applications and financial aid deadlines.