What the newest AI models can do
Latest demos, by model
Stories
Parliament's human rights committee wants a law banning AI that could escape human control.
Ten evidence sessions, witnesses from Meta, Microsoft, Google and Ofcom, and the recommendation that matters is number 200.
Dario Amodei asked the AI labs to slow down. Within nine hours, Altman, Musk and Hassabis agreed.
Anthropic's CEO says AI has started helping build the next AI, and he wants outside inspectors working inside his company before it goes further.
A day after Amodei asked Washington to keep its AI lead over China, Xi offered BRICS an open source AI community.
China's five proposals for the bloc open with AI, and the one thing Xi asks the rest of the world for is a framework everybody writes together.
In brief
- Sam AltmanAltman named the two things OpenAI wants to avoid: losing control of the future to AI, and power over it concentrating in one country, company or lab.
- Sam AltmanSam Altman said OpenAI welcomes a federal framework for frontier AI but won't wait for an anti-trust exemption or legislation, and now writes explicit safety cases before frontier reinforcement learning runs it expects to raise capability.
The current model from each lab
| Lab | Newest model | Origin | Status | Released |
|---|---|---|---|---|
| OpenAI | GPT-6 AstraOpenAI · US · rolling out | US | rolling out | 2026-09-03 |
| Gemini 3.8 FlashGoogle · US · stable | US | stable | 2026-09-02 | |
| Anthropic | Claude Fable 5.1Anthropic · US · generally available | US | generally available | 2026-09-01 |
| DeepSeek | DeepSeek-V4-ProDeepSeek · China · generally available | China | generally available | 2026-08-13 |
| xAI | Grok 4.6xAI · US · generally available | US | generally available | 2026-08-12 |
| Alibaba | Qwen3.8Alibaba · China · open weights | China | open weights | 2026-08-12 |
| Meta | Muse Spark 1.1 · Muse Spark 1.3 shipped 2026-09-02Meta · US · generally available | US | generally available | 2026-07-09 |
| Mistral | Mistral Medium 3.5Mistral · France · generally available | France | generally available | 2026-04-26 |
Checked against each lab's own docs on 2026-09-04. Google's Pro tier lags its Flash tier, and Meta has shipped no Llama 5. Model Watch · release history · compare two
What the flagships cost
OpenAI
$10.00 in, $50.00 out · about $26 a monthAnthropic
$10.00 in, $50.00 out · about $26 a monthxAI
$2.00 in, $6.00 out · about $3.52 a monthPer million tokens, list price, from each lab's own docs. The monthly figure is 700 words in and 500 out, twenty times a day. Put your own numbers in, find one that fits or compare two.
Ready-made prompts
Pick one, change the bits that are yours, paste it into ChatGPT or Claude. Every prompt arrives filled in, so you can see what you're getting before you type a word.
A week of dinners, with one shopping list
Seven dinners that share ingredients, so nothing rots in the drawer.
CookingCatch me up before the new season
You watched it two years ago. Get the plot back in five minutes, no spoilers past where you stopped.
Stories & mangaChase money someone owes me
The message to send today, and what to do when they ignore it.
Writing & memosA study plan for the time I've actually got
Works backwards from the date, and tells you what to drop when you fall behind.
LearningA present for someone you actually know
Tell it what they're really like. Get ideas that aren't a gift guide for their demographic.
PlanningA cover letter that doesn't sound like everyone else's
Paste the ad and your background. Get a short letter that points at real things you've done.
WorkTools that run in your browser
Nothing you paste into these leaves the page. No account, no upload, no key.
Writing Tells
Paste text, get a score and the named signals behind it: em dashes, tricolons, contrast negations, the lexicon.
set up once, then forgetReady-made skills
Install one and Claude reaches for it on its own: meeting memos, flashcards, replies in your own voice.
Claude · ChatGPT · GeminiPrompt Builder
Role, context, task, constraints, examples and output format, rendered the way each model reads best.