Claude Haiku 5.5 costs a tenth of Haiku 4.5 per token and now handles office work the old model barely could. One line matters: past 100,000 tokens, about 110 pages, the whole request bills at five times the rate. Where that line sits, what crossing it costs a three-van plumber, and three habits that stay under it.
Mistral Large 4 claims 59.9 percent on AutomationBench, 657 real office workflows, without saying which scoring it used. The independent boards show something more useful: the best AI does 93 percent of the work and still loses points to one broken business rule. Three habits that follow.
A new study of 2,383 students found an hour with an AI tutor taught as much as an hour with an expert human tutor, and the cheapest model did it for 6.7 cents. The detail that matters for training a new hire is how every tutor started the hour: with the student’s own mistakes.
Gemini 4 Argon tied GPT-6 Astra on intelligence, but the bigger story is what it does when it does not know. Over 100 hard questions Astra gets 13 more right and makes about 19 confident mistakes to Argon’s 7 or 8. Which jobs in a small business favor each, and what to test this week.
From October 30, ChatGPT Pro 200 keeps its $200 price but drops to 10 times the Plus allowance, and the new Pro 500 sells 25 times. That puts every step of OpenAI’s ladder at $20 per multiple, while Claude Max 20x still sells at $10. Plus how GPT-6.1 Sol and Claude Sonnet 5.5 landed on the same price.
Claude Sonnet 5.5 is now the model on Claude’s free plan, and the free app runs it at Medium effort. Anthropic says even that setting beats Sonnet 5’s best score. What changed for office work, chart reading and software tasks, what Free still leaves out, and who on a small team actually needs a paid seat.
On a real month of 880 drafted customer replies, Grok 4.7 costs $3.52, under GPT-6 Sol and less than half of Opus 5.5. The lead is 88 cents, and it rests on a setting most owners never see: Grok 4.7 always reasons before it answers, the default is high, and the thinking bills as output.
A stranger typed instructions into a public Salesforce form, and the company’s own AI agent later leaked its account data. Salesforce has patched SalesBleed. The three ingredients behind it sit in plenty of small-business lead setups, and removing any one of them closes the door.
Meta’s consumer launch for its $1,299.99 VR Glasses never uses the word metaverse. The developer post a day later says VR is expanding, points builders to the Horizon Store and offers $1 million in prizes. Reality Labs still loses about $50 million a day. The metaverse is being built, just not named.
A 41-country study found companies adopting AI grew their headcount, and the new seats went mostly to experienced workers. The junior job loss in the headlines is not statistically significant in the paper itself. What that means for the owner who still has to train the next senior tech.
