Anthropic released Claude Opus 5 on Friday, a model the company says delivers nearly all the intelligence of its top-of-the-line Claude Fable 5 at half the cost — a launch that signals how the AI race is shifting from raw capability to the economics of daily use.

The model, available immediately on all of Anthropic’s platforms, is priced at $5 per million input tokens and $25 per million output tokens, unchanged from its predecessor, Opus 4.8. It becomes the new default model on Claude Max, Anthropic’s premium consumer tier, and the strongest model available on Claude Pro.

The positioning is deliberate. Anthropic is not claiming Opus 5 is its smartest model — that distinction still belongs to Fable 5, and rival systems retain an edge in certain domains. Instead, the company is making a subtler argument that may matter more to enterprise buyers: that the most economically important AI work happens in a middle band of difficulty, where near-frontier intelligence delivered efficiently and cheaply beats frontier intelligence delivered expensively.

“Opus 5 as your daily driver, the model you hand complex work to and review when it’s done,” an Anthropic spokesperson said in an interview with VentureBeat, describing how the company’s lineup now stratifies. “Fable 5 for your most ambitious work, the days-long autonomous projects nothing could take on before… Sonnet 5 for work you run at scale, where speed and cost per call decide what ships. Haiku 4.5 for subagents and instant answers.”

How Claude Opus 5 benchmark results stack up against Fable 5 and rival AI models

On paper, the results are striking. Anthropic says Opus 5 sets new state-of-the-art marks on coding and knowledge-work evaluations including Frontier-Bench and GDPval-AA. On Frontier-Bench v0.1, an agentic terminal coding benchmark, Opus 5 scores 43.3 percent — more than double Opus 4.8’s 18.7 percent and well ahead of Fable 5’s 33.7 percent — at a lower cost per task, according to the company. On ARC-AGI 3, an evaluation of novel problem-solving, Anthropic reports Opus 5 scored three times as high as the next best model. On OSWorld 2.0, a computer-use benchmark, the company says the model surpasses Fable 5’s best result at just over a third of the cost.

The numbers come with honest caveats that are themselves notable in an industry prone to superlatives. Anthropic acknowledges Opus 5 remains behind Mythos 5, a competing model, on cybersecurity tasks and biology research, and an OpenAI-family model still leads on one agentic coding benchmark.

The more revealing caveat came from Anthropic itself, when asked where Opus 5 still falls short of Fable 5. The spokesperson’s answer amounted to a candid admission about what benchmarks do and don’t capture.

“The evals where Opus 5 wins are bounded tasks with a specific outcome, which is where it’s strongest. What those evals don’t measure is duration,” the spokesperson told VentureBeat. “One way to put it: Opus 5 is the best tool for the jobs benchmarks can see, and Fable 5 is what you reach for when the job outruns the benchmark.”

Fable 5, by contrast, “is for the longest, most autonomous jobs, where the model has to stay coherent across many connected steps over hours or days with dense source material,” the spokesperson said, advising customers to “run both on a representative workload, one bounded task and one long-horizon job.” That framing — bounded tasks versus long-horizon autonomy — may become the defining axis of model differentiation in 2026, as benchmarks saturate and the hardest remaining problems involve sustained, multi-day agentic work rather than discrete puzzles.

Why token efficiency is becoming the real battleground for enterprise AI spending

Threaded through the launch is a theme Anthropic clearly wants buyers to absorb: Opus 5 doesn’t just score well, it scores well per dollar. The model ships with an adjustable “effort” setting that lets customers trade intelligence for speed and token savings, and Anthropic’s charts emphasize performance at a given cost rather than peak performance alone.

Early customers echoed the point with unusual specificity. Harvey, the legal AI company, said Opus 5 achieved similar performance to Opus 4.8’s maximum-reasoning mode “while generating 26% fewer tokens on average,” according to Niko Grupen, its head of applied research. Richard Pham of Fundamental Research Lab said that on hard financial-modeling tasks, the model averaged nine percentage points higher accuracy “while using roughly one-third fewer turns and tool calls and 60% less time.”

Wade Foster, chief executive of Zapier, said Opus 5 topped his company’s AutomationBench leaderboard “without spending more tokens than prior Claude models,” running a full churn-prevention workflow from start to finish. “Previous models didn’t pass; Opus 5 hit 100%,” he said. Scott Wu, chief executive of Cognition, the company behind the Devin coding agent, said that on FrontierCode 1.1, “Claude Opus 5 approaches Fable-level performance at half the cost,” with particular strength in debugging and root-cause analysis.

The efficiency emphasis reflects commercial reality. Enterprise AI spending is no longer experimental, and inference costs — the price of actually running these models at scale — have become a board-level line item. 

Anthropic’s business skews heavily toward API and enterprise usage; according to a February 2026 analysis by Contrary Research, Claude held roughly 40 percent of the enterprise large language model market by usage as of late 2025, and Claude Code alone had reached about $1 billion in annualized revenue. For a company whose customers pay by the token, a model that does more with fewer tokens is not a nice-to-have. It is the product.

Self-verifying AI agents and what they mean for the hidden costs of automation

Beyond the numbers, Anthropic is selling a behavioral story: that Opus 5 verifies its work and iterates until it succeeds. The company offered several examples from testing that read like small parables of machine stubbornness.

In one Frontier-Bench task, the model was asked to reconstruct a machine part as a 3D CAD model from a drawing it was intentionally given no way to view. Rather than fail, Anthropic says, Opus 5 wrote its own computer vision pipeline to extract the geometry from raw pixels — and did so repeatedly, while no competing model solved the task in five attempts. In another case, given a real bug in a popular open-source package manager, the model found the root cause and fixed an edge case the community’s own patch had missed; a competing model patched only the symptom and declared victory. An engineer at a trading firm, the company says, used Opus 5 to build a market data feed for a new exchange in a single session and, finding no live feed to validate against, watched the model build its own test harness to check its parsing code.

Customers described similar behavior in the wild. Cristian Rivera, a staff software engineer at Stripe, said he gave the model “a chief-of-staff role over my dev environments” for a weekend: “it built its own monitor, drove each box, and pulled me in only for the judgment calls.”

This is the capability enterprises actually care about, and it is worth dwelling on why. The gap between a model that produces plausible output and one that verifies its output is the gap between a demo and a deployable system. Most of the hidden cost of enterprise AI today is human review — engineers checking the machine’s work. A model that reliably checks its own work compresses that cost, which is precisely why customers keep citing fewer turns, fewer passes, and less time rather than higher raw scores.

Inside Anthropic’s safety strategy: capability gaps, classifiers, and model fallbacks

The launch also showcases Anthropic’s increasingly intricate approach to safety — one that now involves deliberately not teaching its models certain skills. The company says its automated behavioral audit found Opus 5 to be its most aligned model to date, scoring 2.3 on overall misaligned behavior, lower than Opus 4.8, Sonnet 5, or Fable 5, with the lowest rates of deceptive behavior and the least susceptibility to being tricked into misuse.

On the capability side, Anthropic says it intentionally avoided training Opus 5 on cyber tasks, as it did with Opus 4.8. The model improved on them anyway — a side effect of general capability gains — and now nearly matches Mythos 5 at finding software vulnerabilities. But it remains far behind at exploiting them: on Anthropic’s OSS-Fuzz evaluation, Opus 5 identified vulnerabilities at a 79.4 percent rate, close to Mythos 5’s 80 percent, but succeeded at developing exploits in only 4 challenges versus Mythos 5’s 13. That asymmetry — strong at defense-relevant discovery, weak at offense-relevant exploitation — appears to be by design, and the safeguards follow the same logic. Anthropic expects Opus 5’s cyber classifiers to intervene about 85 percent less often than Fable 5’s.

When a classifier does trigger, requests in Claude.ai, Claude Code, and Claude Cowork fall back to Opus 4.8 by default — raising an obvious question: if a request is too risky for one model, why is it acceptable for another? “The model it falls back to has lower capability levels making the risk of harmful use lower as well,” the spokesperson said, adding that “there is a message that lets the user know when this occurs and is visible in the chat.”

The logic is defensible, but it reveals how AI safety actually works in 2026: risk is not a property of the question alone, but of the question multiplied by the capability of the system answering it. On biology, the calculus runs the other way. Opus 5 is now Anthropic’s most capable generally available model for scientific research — scoring 10.2 percentage points higher than Opus 4.8 on the company’s internal chemistry benchmark — though the spokesperson acknowledged that “Mythos 5 remains the stronger model for long-horizon, open-ended work like autonomous drug design campaigns.”

The business stakes behind the launch: a $380 billion valuation and massive compute bets

The launch lands at a moment of extraordinary commercial momentum — and extraordinary obligations — for Anthropic. Reuters reported in February that the company was valued at roughly $380 billion in its latest funding round, following a period in which, per Contrary Research’s analysis, its annualized revenue climbed from about $1 billion at the end of 2024 to a projected $9 billion by the end of 2025, with internal targets reportedly reaching $20 to $26 billion for 2026. Those targets are underwritten by enormous infrastructure commitments, including a reported $30 billion Azure compute deal alongside arrangements with Google Cloud and Nvidia — spending that only pencils out if enterprises keep expanding usage.

That is the context in which Opus 5’s pricing strategy makes sense. Holding the price at Opus 4.8 levels while roughly doubling performance on key agentic benchmarks is effectively a steep price cut per unit of capability, designed to widen the funnel of workloads that are economical to automate. Every task that was marginal at Opus 4.8’s cost-per-success becomes viable at Opus 5’s — and every viable task is recurring token revenue.

The regulatory backdrop has grown more complex as well. A U.S. judge gave final approval this week to Anthropic’s $1.5 billion copyright settlement with book authors, Reuters reported, closing a chapter of litigation over the company’s early training data. And in June, Reuters, citing Axios, reported that the U.S. government had moved to block foreign access to Anthropic’s most advanced models — a reminder that frontier AI is now entangled with export policy in ways that shape which customers can buy what.

Also shipping Friday: a Fast mode running at roughly 2.5 times default speed at twice the base price, automatic fallback routing on the API, and mid-conversation tool changes that no longer invalidate the prompt cache — a small feature that agent developers may appreciate more than any benchmark. Consistent with prior Opus models, Opus 5 carries no data retention requirements for general access, a point the spokesperson flagged unprompted for customers with “a hard zero data retention requirement.” Developers can access the model as claude-opus-5 on the Claude API starting today.

Two questions will determine whether the bet pays off: whether Opus 5’s efficiency claims survive contact with production workloads at scale, and whether enterprises embrace a world where safety classifiers, not users, sometimes decide which model answers. But the deeper message of Friday’s launch is that the AI industry’s center of gravity has moved. For three years, the labs competed on what their best model could do on its best day. With Opus 5, Anthropic is competing on something less glamorous and far more lucrative: what a very good model can do every day, for half the price. In a market where the frontier keeps moving, Anthropic is wagering that the real fortune lies just behind it.

Some of the best series on Apple TV have earned great reviews, yet none of them get talked about nearly as much as they deserve. That’s why I have included three different genres in this weekend’s binge list, which is based on a bestselling sci-fi novel, a beloved wine manga, and a true-crime memoir. The best part is that each one of these shows manages to outdo its source material’s reputation on the screen.

We also have guides to the best new movies to stream, the best movies on Netflix, the best movies on Hulu, the best free movies, and the best movies on Amazon Prime Video.

Dark Matter (2024-present)

Genre: Sci-fi, Drama, Thriller
IMDB rating: 7.6/10
Rotten Tomatoes: 81%

Physicist Jason Dessen (Joel Edgerton) is walking home one night in Chicago when he’s abducted and wakes up in an entirely different version of his own life, one where he made a very different choice at a critical fork in the road years earlier. Desperate to get back to his real wife Daniela (Jennifer Connelly) and son Charlie (Oakes Fegley), Jason has to navigate a disorienting maze of alternate realities, all while facing off against the most dangerous version of himself imaginable.

Based on Blake Crouch’s bestselling novel, the show blends a mind-bending sci-fi premise with emotions at its core. Edgerton carries the weight of playing multiple versions of the same man without ever losing the thread of who Jason really is. Season two premieres August 28, so this is the perfect weekend to catch up before it returns.

Stream Dark Matter on Apple TV.

Drops of God (2023-2026)

Genre: Drama, Mystery
IMDB rating: 8.0/10
Rotten Tomatoes: 96%

When legendary wine critic Alexandre Leger dies, his estranged daughter Camille (Fleur Geffrier) learns she stands to inherit his extraordinary collection, but only if she can outduel his star protege Issei (Tomohisa Yamashita) in a brutal series of blind wine-tasting tests. The catch is that Camille has spent years physically unable to drink alcohol due to a traumatic childhood memory she’s never fully confronted.

Spanning France and Japan, the show turns the notoriously stuffy world of wine expertise into something genuinely tense and cinematic. It earned high score on both Rotten Tomatoes and IMDB for its first season, and season two pushes both leads into an even bigger mystery, tracking down the origin of the greatest wine ever made. It’s a hidden gem that rewards patience with real emotional payoff.

Stream Drops of God on Apple TV.

Black Bird (2022)

Genre: Crime, Drama, Biography
IMDB rating: 8.1/10
Rotten Tomatoes: 98%

Jimmy Keene (Taron Egerton) is a former high school football star turned convicted drug dealer facing a full decade behind bars, until the FBI offers him an impossible deal. Transfer to a maximum-security prison for the criminally insane, befriend suspected serial killer Larry Hall (Paul Walter Hauser), and get him to confess to the murders of as many as eighteen young women before his appeal goes through.

Written by novelist Dennis Lehane, this six-episode limited series turns a real-life story into a slow-building psychological standoff between two very different kinds of dangerous men. Hauser’s performance as Hall is surreal, and the late Ray Liotta delivers one of his final on-screen performances as Jimmy’s ailing father. It’s tense, true, and easy to finish in a single weekend.

Stream Black Bird on Apple TV.

The phone rings. There’s a sense of dread. Friend or foe? Scam or no? NordVPN — known more for encrypting internet traffic and hiding IP addresses — has launched an Android app named HangUp that screens incoming calls for potential scammers and telemarketers. The app is free to everyone. You don’t need a paid NordVPN account.

HangUp uses call protection technology from Hiya, which has its own mobile app to block spam and fraud calls and detect AI deepfakes. HangUp also can scan incoming text messages by incorporating NordVPN’s Message Protection, which the company launched in June. HangUp app customers can choose to use only call screening if they wish.

Flagging calls might be a new feature for a VPN that CNET deems the fastest, but scam screening is already available via many cellphone service providers as well as devices themselves. AT&T, T-Mobile, Verizon, iPhones, Google Pixel and Samsung phones can identify and block unwanted calls, as can major landline services.

HangUp is currently available only for Android and can be downloaded in the Google Play Store. After downloading the app, you’ll need to sign in with your NordVPN account or create a free one on its website. Finally, you’ll need to give permission to the app to identify caller ID and spam.

NordVPN premium customers already have call and message protection built into the app.

Worldwide problem

Scam calls aren’t just annoying — they defraud people out of billions of dollars annually. It’s a far cry from the quaint beginning of telemarketing in 1957.

Nearly 80% of Americans got at least one scam call in the past year, and 40% report feeling unsafe because of scam calls, according to a NordVPN survey. Hiya research found that Americans receive an average of 10 unwanted calls per week and that 30 million people lost money — at an average of $682 apiece — to a phone scam in the past year.

See also: Scam Calls Have Gotten Out of Control. This One Strategy Actually Helps

NordVPN’s survey said the most common scam in the US is fake telemarketers or sales agents (43%), followed by fake healthcare providers (20%) and fake government or tax officials (18%).

“What these numbers reflect is a deeper anxiety about not being able to trust a call,” NordVPN Product Director Domininkas Virbickas said in a statement.

How HangUp works

When a call comes in, HangUp checks the number and warns you if it could be a scam, spam or telemarketing attempt before you answer the call. The app is able to do this by using caller reputation data that Hiya uses for its own service.

You can’t block individual phone numbers with the app, but instead, you can block categories — for instance if you want to stop all calls that are categorized as telemarketers.

You can also employ HangUp to scan incoming text messages. The app says it can detect scams, fraud, malicious messages and deceptive content. The app says it “may analyze message content, including links, sender information, and other relevant signals, to assess potential risks.”

If you’re worried that this is another way of getting your data, NordVPN says it will not record or listen to calls, nor will it store any text message content.

NordVPN has another disclaimer about HangUp. It says the app is only an “informational assistance tool” and can’t guarantee that it will be able to identify and block all unwanted calls. HangUp users can’t “rely exclusively” on the app and are “solely responsible” for calls and texts that they get.

So be sure you know the warning signs of scam calls and the steps you can take. The Federal Communications Commission lists them out, including not answering calls from numbers you don’t know and hanging up immediately if someone asks for money or personal information.

Ever since being admittedly fascinated by the Cambridge coffee webcam from the 1990s, I’ve written about VPNs, the NFL, smartphones, living wages, over/unders and everything in between.

See full bio

[This is a press release by the National Technology & Innovation Sandbox (NTIS).]

The National Technology and Innovation Sandbox (NTIS) has been named a Champion at the World Summit on the Information Society (WSIS) Prizes 2026 in Geneva, under Category 14: Action Line 7: E-science.

An annual global awards programme, WSIS Prizes celebrates outstanding digital and ICT-based initiatives that advance the UN Sustainable Development Goals (SDGs). The contest is organised by the International Telecommunication Union (ITU) and highlight success stories from government, private sectors, civil society, and academia.

Category 14: Action Line 7: E-science, focuses on leveraging information and communication technologies (ICTs) to advance scientific research, facilitate open data sharing, and foster global collaboration to achieve SDGs.

This win reinforces Malaysia’s position as a regional innovation leader and marks the nation’s strategic shift towards a dynamic “living lab” model, bridging the gap between innovation and market readiness.

Spearheaded by the Ministry of Science, Technology and Innovation (MOSTI) with MRANTI as lead secretariat, NTIS has evolved since 2020 from a pilot initiative into a cornerstone of Malaysia’s innovation ecosystem.

The initiative provides a structured platform for researchers and startups to validate high-growth technologies, ensuring they are regulatory-compliant and commercially viable. Building upon the foundational achievements of the 12th Malaysia Plan (RMK-12), the NTIS programme continues to accelerate its momentum well into the 13th Malaysia Plan (RMK-13).

Since 2020, NTIS has facilitated over RM171.8 million in commercialisation value, approved RM147 million in funding to 147 projects, catalysed 388 high-income jobs, generated RM122.8 million in export value, and secured 39 new Intellectual Properties.

Reflecting on the achievement, YBhg. Dato’ Ts. Dr. Nagulendran Kangayatkarasu, Secretary General of MOSTI said, “This achievement reflects our commitment to elevating the competitiveness of homegrown technology companies. NTIS will continue to serve as a vital platform, addressing challenges faced by local startups and increasing the volume of commercialised innovations, thereby reducing dependency on foreign technology and strengthening local capabilities.”

Echoing this sentiment is Mr. Ashran Dato’ Ghazi, CEO of MRANTI, “Recognition as a WSIS Champion validates Malaysia’s progress through NTIS. We have demonstrated that with the right ecosystem, regulatory support, and partnerships, promising technologies can move from research to real-world deployment. This success belongs to the entire ecosystem of innovators, partners, and agencies that made this model possible.”

The programme’s success is underpinned by a collaborative multi-agency framework led by MRANTI as the lead secretariat, together with co-lead MTDC, and other secretariats MIMOS and Futurise. 

Together, these organisations play complementary roles in strengthening regulatory readiness, technology validation, commercialisation and ecosystem coordination, enabling Malaysian innovations to progress more effectively from development to market.

Building on this strong foundation, NTIS is now entering its next phase through the Global Testbed & Sandbox Alliance (GTSA), expanding its focus beyond national deployment towards greater international collaboration.

The initiative strengthens cross-border partnerships and regulatory harmonisation, enabling Malaysian innovators to scale solutions across ASEAN and global markets while engaging directly with UN bodies, international policymakers and leading innovation ecosystems.

• Learn more about NTIS Programme here. 
• Read articles we’ve written about Malaysian business here. 

Featured Image Credit: MRANTI

Jamie wanted a machine that would make her write. Not a laptop full of open tabs and alerts. Not a phone that vibrates every few minutes. She wanted something that felt like it belonged on a desk from another decade, yet still saved every word without paper or ink ribbons. So she spent weeks designing and building one herself. The result sits somewhere between a classic typewriter and a modern writing deck, finished in green paint with rounded keys and a screen that rises on command.

She began with a keyboard, which was a small mechanical board sporting a built-in circuit and switches. The original square keycaps were removed and replaced with ones she produced herself; however, these ones had a rounded top and looked more like the keys on an old typewriter. Labeling them was simple; simply color the tops with some paint markers, then run them through a laser engraver while the paint is still wet, and the laser will burn the ink into the plastic, ensuring it lasts. Any extra paint easily wipes away, leaving some lovely sharp letters.

XTEINK X4 E-Book Reader (Developer Edition), 4.3″ Pocket E-Ink eReader, Physical Page-Turn Buttons…

• All overseas models purchased through official channels – both past sold and current for sale– have not and do not restrict users from installing…
• Pocket-Size Mini eReader for Reading Anywhere: Ultra-light at just 0.23 inch and only 2.72 oz, Xteink X4 is designed for true portability. Slip it…
• 4.3″ Paper-Like E-Ink Display: The 4.3-inch E-Ink screen delivers a natural paper-like reading experience that’s gentle on the eyes. Enjoy clear…

The body of the machine took a little longer to complete, as Jamie created a true full typewriter shell in CAD and began working on the shape and dimensions until it felt just perfect. Then she printed it in sections. Once they were out of the printer, she connected them all together with ABS and acetone, a unique compound that chemically welds the plastic together. Then she slathered on a second coat of the mix to fill in the lines left by the printer, followed by some sanding, priming, and one coat of green paint, and it began to resemble a true typewriter rather than a kitbash of printed pieces.

The Raspberry Pi Zero 2 W is powered by a rechargeable lithium battery, while the display is an e-paper screen from Waveshare, which is quite sleek to. say the least. The software is simply a customized version of the open Zerowriter project, which is used by the majority of other distraction-free writing machines. Jamie poked at the code to get the cursor moving with the arrow keys and copy and paste working, which just transforms it from a pure draftsman’s tool to something you can use for actual writing on the fly.

The two mechanical bits give this machine its typewriter feel. One side has a handwheel that can be turned to raise and lower the e-ink screen, similar to how a sheet of paper would pop up from a real typewriter. It’s quite pleasant to use. On the other side, there is a rotary encoder wheel, which you may use to open the menu and scroll through your content. Right next to that is a small seven-segment display that shows how many words you’ve written thus far. When you press a key, the word count refreshes.

The machine does exactly what she intended to build. It sits on a desk, accepts keystrokes, displays clear text on a paper-like surface, maintains a running word count, and asks for almost nothing else. For those who prefer the ritual of a typewriter without the ribbons, carbon paper, or permanent commitment of ink on a page, the green writing deck provides that experience while still saving every document.
[Source]