This Week’s AI Price Cuts Give Workers Leverage, Not Just Savings

Claude Opus 5, released by Anthropic on July 24, 2026, costs roughly half of what Fable did and adds a low, medium and high effort dial you control per task, while a free, open-weight rival from Moonshot AI arrived three days later, giving professionals two new ways to control their own AI costs.

What Did Anthropic Change?

Anthropic’s July 24, 2026 announcement says Claude Opus 5 now costs half of what Fable charged for the same tier, positioning it as the new default model. The company says Opus 5 roughly doubles Claude Opus 4.8’s score on its coding benchmark at that same price, according to the same announcement. The bigger change for daily use is the effort dial: a low, medium and high setting that lets you choose how much reasoning the model spends on a task, per Anthropic. Set it to low for a quick email draft and high for a contract review, and the price moves with it. That doubled benchmark score is a vendor claim from Anthropic’s own testing, not an independent audit, so treat it as a marketing figure until outside labs replicate it.

Why Should You Care About the Dial?

The effort dial turns AI spending into a decision you make per task instead of a fixed subscription cost, according to Anthropic’s Opus 5 announcement. Before this release, most professionals picked one model tier and paid that rate whether they were writing a one-line reply or debugging a production outage. Anthropic now frames Opus 5 as usable across coding, research and document work with cost and quality traded off in the same session, per the same announcement. For anyone billing clients or managing a team budget, that means you can defend a high-effort run with a paper trail showing the setting, the task and the output. It also means a manager comparing your AI spend to a colleague’s can no longer assume flat cost per person. The dial pushes the judgment call about how much a task is worth onto the person doing it, a new kind of accountability alongside the discount.

Did AI Cause the Layoffs?

Monday.com cut 600 jobs on July 25, 2026, and cited an AI-driven strategy change as the reason, according to TechCrunch’s running list of 2026 layoffs where employers named AI. That is the company’s own explanation, not an outside audit of which tasks AI replaced, so it does not prove AI systems handled the eliminated roles rather than plain cost-cutting dressed in AI language. The same TechCrunch tracker lists Monday.com alongside other software companies restructuring in 2026 and naming AI as the driver. For a working professional, the honest reading is narrower than the headline: a company choosing to frame layoffs around AI tells you what story leadership wants investors to hear, and it is worth watching whether the specific roles cut were the ones most exposed to AI tools before you treat one case as proof of a trend.

Why Would You Use a Free Model?

Moonshot AI released Kimi K3 on July 27, 2026, calling it the largest open-weight model yet, and independent researcher Nathan Lambert reported it performs close to closed frontier systems on his site Interconnects. Open-weight means anyone can download the model and run it on their own hardware instead of paying per token to a lab like Anthropic or OpenAI, a point Lambert makes directly in his July 27, 2026 write-up. That changes the calculation for teams without a corporate API budget: a solo developer, a small agency or a student can now self-host something close to a frontier coding assistant for the cost of the hardware alone. Lambert frames this as an escalation in the open-weight race, per Interconnects, and it is worth remembering that “close to” is not “equal to.” Kimi K3 likely still trails the newest closed models on the hardest benchmarks, so treat it as a serious budget option, not an automatic replacement for Opus 5 on your toughest tasks.

Four Moves, Ranked by Urgency

Rank these four moves by how fast the underlying advantage closes, starting with the one that expires first.

1. Test the effort dial on a real task this week. Run the same document or code review at low, medium and high effort in Claude Opus 5 and compare the output against the price difference, per Anthropic’s announcement, before your manager sets a default for the whole team.
2. Download Kimi K3 and benchmark it against your current tool. Moonshot AI’s model costs nothing to run, according to Interconnects, so there is no cost to finding out where it falls short before you need it under deadline pressure.
3. Ask what your employer would cite if it cut your role. TechCrunch’s 2026 layoffs list shows companies increasingly naming AI in public statements, so know which of your tasks look automatable on paper even if they are not in practice.
4. Write down how you use the effort dial for billing or reviews. A record of when you chose high effort and why becomes evidence of judgment, the harder thing for a model or a layoff announcement to replace.

What Should You Do Right Now?

Right now, pair the cheaper frontier model with the free open-weight one instead of picking a side. Use Claude Opus 5’s effort dial for tasks where a wrong answer is expensive, such as a client-facing document or production code, since Anthropic’s own testing claims a coding-benchmark gain at unchanged high-effort pricing. Use Kimi K3 for lower-stakes drafting, internal notes or experimentation where per-token cost genuinely changes your monthly bill, following the approach Lambert describes at Interconnects. Neither choice protects you from a layoff framed around AI, which is why the Monday.com case covered by TechCrunch is worth reading in full rather than as a headline. The professionals who come out ahead this year will be the ones who can show, concretely, which parts of their job an effort dial or an open model still cannot do.

You are a pragmatic AI cost advisor helping me decide how to use Claude Opus 5's effort dial for my actual work. I will describe a task I need done (for example: drafting a client email, reviewing a section of code, or summarizing a report). Based on what I describe, recommend whether I should run it at low, medium or high effort, and explain in plain language what I am trading off in quality versus cost at each level. Ask me what the task is, who will see the output, and how much a mistake would cost before you answer. If you are not certain how the effort levels differ for my type of task, say so instead of guessing.

Act as a skeptical career strategist, not a cheerleader. Help me list the tasks in my current job that a general-purpose AI model could plausibly do today, and separate them from tasks that still require judgment only I can provide. Ask me to describe my role, my main weekly tasks, and who reviews my work before you answer. For each task, tell me honestly whether it looks automatable, partly automatable, or not automatable with today's tools, and explain your reasoning rather than just labeling it. Do not reassure me if the honest answer is uncomfortable.

You are helping me run a fair, side-by-side comparison between a free open-weight model like Kimi K3 and my current paid AI subscription, for one specific task I do regularly. Ask me what the task is, what tool I currently use, and what a good result looks like to me before giving any advice. Then design a short test I can run myself, with the same prompt on both tools, and tell me what to look for in the outputs to judge accuracy, tone and reliability, not just speed. Flag any part of my task where a mistake would be costly enough that I should not trust a free or unfamiliar model yet.

Act as a blunt hiring manager who has read ten thousand resumes, not a career coach. I will paste my full resume and the job description I want. Rewrite the whole resume for that role, section by section, in this order: summary, experience, skills, education. Rules: every experience line leads with impact, not duty. Use bracketed placeholders like [8 percent] for any number I did not give you, and list at the end every placeholder I need to replace with a real figure. Keep it to one page of text. Plain formatting only, no tables or columns, so screening software can parse it. After the rewrite, tell me the three weakest claims that need evidence before I send this anywhere. My resume: [paste resume]. The role: [paste job description].