News

• Huntress uncovers phishing campaign abusing Meta’s business account email infrastructure and impersonating the Meta Agency Partner Program
• Victims were tricked into handing over credentials, which attackers exfiltrated to Telegram for account takeover, scam ads, and targeted phishing
• Meta has since added guardrails that killed the campaign; Huntress published IoCs to help organizations detect related activity

Hackers are abusing one, and impersonating another legitimate Meta service, to try and steal login credentials for people’s business accounts with the company.

Meta, the owner of Facebook, Instagram, WhatsApp, and more, allows businesses to set up separate accounts and talk to each other. The emails sent from one to another pass through the company’s infrastructure, meaning they are shown as coming from Meta itself.

However up until recently, hackers were abusing this fact to send phishing emails that landed directly into their victims’ inboxes, security researchers Huntress explained. The company even tried to curb this by hardcoding a disclaimer that email senders are not part, or affiliated with, Meta, but crooks found creative ways around this, as well.

The phishing emails redirected victims to landing pages outside Meta’s ecosystem. These pages were designed to mimic Meta Agency Partner Program, a legitimate initiative that connects businesses with professionals in social media management work.

Those who would fail to see the ruse would end up trying to log into their accounts, instead just sharing their login credentials with the attackers. The secrets would get exfiltrated to a Telegram account under the threat actors’ control, which they could later use for different things, from phishing to malvertising.

“Threat actors can leverage Meta business accounts to spend the victim's money on malicious or scam advertising, or they can take over the account entirely, changing the recovery methods and password, and leverage the account to transmit more targeted attacks at the business' customers or social media followers.” the researchers explained.

Over the last couple of months, the campaign evolved and changed, using different lures and mechanics, but keeping the same end goal. However, Meta has effectively killed it by adding additional guardrails that now make it impossible to run.

Netflix has proved that it has a knack for producing solid original content, from Stranger Things and Squid Game to Bridgerton, but now the platform is again diversifying into new content areas — and its reputation as one of the best streaming services is starting to slip for me.

Starting on August 3, the platform will be adding video content from a number of popular online outlets including Buzzfeed, Conde Nast, and Variety, spanning topics such as lifestyle, food and drink, fashion, and entertainment. Basically, all the video series you can already find on YouTube will be coming to Netflix.

Netflix says subscribers will be able to stream “some of [their] favorite videos from around the Internet without having to leave Netflix,” as the service puts it — with videos ranging from short two-minute watches, to in-depth 20-minute content. The videos will be available to viewers in the US, Canada, the UK, Ireland, Australia, and New Zealand first.

When the new content lands on the platform, you’ll have another way to watch popular internet video series like Variety’s Lie Detector Test and Architectural Digest’s Walking Tour without having to leave the Netflix app.

Netflix clearly wants to diversify beyond its usual film and TV show offerings. Recently, the service announced its Clips short-form video feed, which is clearly borrowed from TikTok and Instagram Reels. It’s also investing in video podcasts, looking to capture a slice of a market that’s already catered to by Spotify and other services.

I understand that viewing habits are shifting, but Netflix’s attempts at moving into new content feels like a Katy Perry album release, with the streaming giant trying to chase trends for the sake of keeping up appearances, when it should be focussing its time and resources on producing quality original content.

That brings me to the ‘two season problem’.

Netflix has an awful habit of pulling its original shows, even if subscribers love them, after just one or two seasons. The platform just canceled The Boroughs after one season despite it receiving rave reviews — and Stranger Things creators The Duffer Brothers were attached to the project, so a second season was widely expected.

According to the platform, this decision was taken due to production costs (sci-fi is expensive to produce), but somehow Netflix still manages to find enough money to pump into reality shows that receive nowhere near as much critical acclaim as its axed shows get, and which, I’d argue no one really wants.

Although spin-off show Squid Game: The Challenge climbed the Netflix charts, the roaring success of the original TV series helped it massively. But, as our streaming writer Jasmine Valentine says in her two-star review, the second season was “nowhere near as good as its first”. Between this and the pure cringe-fest that was Floor is Lava, you’d think Netflix would’ve learned its lesson — apparently not.

Instead, it’s continued to churn out one reality show after the next. I couldn’t make it past the first 20 minutes of Age of Attraction, and now my favorite childhood movie is about to have the magic sucked out of it thanks to a new Wonka-themed reality show.

I’m not the only one who thinks this, my colleague and streaming writer Lucy Buglass is on the same page, telling me “Netflix’s decision to host content from other platforms has been met with criticism, and there’s a good reason for that. Many beloved Netflix shows have been axed by the streaming service, and in 2024 Netflix canceled these 9 shows despite many of them performing well with both critics and audiences.

“When you compare Netflix to other streaming services, it’s shocking how quick they are to abandon their shows. Competitor Apple TV has a lot of faith in its original programming from sci-fi hit Severance, to horror comedy Widow’s Bay, and the moreish drama Margo’s Got Money Troubles. All of these shows have been renewed for new seasons in the future, which is welcome news for loyal fans.

“I’m concerned that Netflix keeps neglecting some great-quality shows in favor of hosting content we can already get elsewhere. Time and time again, TV fans are not getting a satisfying finale because storylines are cut off abruptly after being axed. This only discourages people from watching new shows if they’re going to be left on a permanent cliffhanger.”

Competing against YouTube is going to be quite a challenge for Netflix, and it makes me wonder if it'll be successful in luring more viewers in its direction. But for the time being, I wish it would just focus on quality of quantity.

• Call of Duty: Black Ops and Black Ops 2 were just ported to PlayStation 4 and PlayStation 5
• Aside from an upgraded picture quality, they seem otherwise identical to the original versions
• Some fans are convinced that they just don't feel right, however

Yesterday, Activision treated us to the surprise release of not one, but two Call of Duty ports. You can now buy the original Black Ops and its acclaimed follow-up, Black Ops 2, on PlayStation 4 and PlayStation 5 for the very first time.

They upgrade the resolution of both games on these newer consoles (a substantial step up from the 608p native picture quality of the backwards compatible versions available on Xbox) but otherwise seem almost identical to their original incarnations.

Even some big features that many fans thought would be cut, like Black Ops 2's infamous emblem creator, are present and accounted for.

I picked up copies of the two games immediately with my own cash (there's currently an irresistible 50% discount on them if you have PS Plus) and spent the best part of yesterday evening playing on my PS5. Aside from the crisper picture quality, they're just as I remember.

I replayed the original versions of both games pretty recently too, via second-hand physical copies that I bought for my Xbox Series X to refresh my knowledge of both games ahead of the Black Ops 7 launch — this latest entry is a Black Ops 2 sequel, after all.

Some fans aren't as impressed, though, and argue that the games just don't feel right.

"Black Ops 1 PS5 has a serious controller input lag issue that needs to be adressed [sic] and fixed" argued one poster on r/callofduty. "It surely wasn’t this bad on PS3".

"Sounds like the original launch" another fan replied, with someone else chiming in: "The golden days" and a tongue-in-cheek sobbing emoji.

"BO1 always felt hard to aim in even in the old days compared to BO2. Pretty sure this is normal" added someone else.

Another post similarly reckons that "The responsiveness is terrible. I distinctly remember the original Black Ops 1 feeling way crisper than this". The top comment agrees, complaining that "I’m losing every gun fight because my gun doesn’t fire in time".

Although everything feels fine to me, there's definitely a chance that those playing specific resolutions, on specific TVs, or with features like VRR enabled might be having a much worse experience. I expect some patches will be released in the coming weeks if these claims hold water.

For the last three years I’ve been imploring anyone who will listen to assuage the AI Fear Economy and the AI Doomonomics messaging that dominates our current AI news cycle.

The sleepwalkers among us and the ones that stand to financially benefit from the FOBO (Fear of Becoming Obsolete) will continue the walking dead path but many experts are finally disagreeing with the narrative that I have been spearheading for over three years now.

I’ve spent countless hours proving AIs limitations in numerous podcasts, interviews and keynotes along with my “AI Exposed” series. I really cannot find the “off” button for exposing the truth.

If you use AI every day, you know the foundational flaws of the transformer-based generative LLM kingdom.

You see the mistakes, errors, forgetfulness, hallucinations, sycophancy, recency bias, kindergarten math fails, regression to the mean, the waiter problem, no real memory, no real world experience, no determinism, no thinking, no cognition, no sense of time, no real context (environmental, cognitive bias, natural behaviors, awareness, etc.), no uniqueness, no native ethics, no agency, no intelligence (really), no accountability, no “strawberry”. Need more?

AI was never built to do the things the industry is asking it to do. Current LLMs are word predictors. They are the trillion dollar equivalent to an autocomplete on acid. Nothing more.

Workforce Extinction is what the economy is selling but Human+AI empowerment is what will be our final state. We just have to get through the current AI dystopia superbubble burst first unfortunately - which might take a bit more time than we would all like. Remember ATMs? Did they replace bank tellers? No. Remember the corner grocery store? Did Amazon eliminate them after buying Whole Foods in 2017, almost 10 years ago? No.

Human: “This is my 6th meeting today. I’m exhausted.”

AI: “I’ve got your back.”

The bane of our existence, I know certainly mine. I am not talking about the nice small talk or the social and connective tissue that creates strong bonds and trust and loyalty (all of which AI cannot replicate today in the slightest), but the mechanical and mundane meetings that litter our week, every week.

Look at your calendar right now and identify next week’s meetings. Are they “ideation” or “decision making” meetings? The dreaded “Sync” meetings? Or the “Stand-ups”? All of these can be augmented if not completely replaced by AI in a powerful synchronicity that solves the “cold start” problem, immediately empowering you to make sound decisions collectively and collaboratively without you being physically present. After all, ideation is all about tapping into the potentiality of an outcome given all the relevant and contextual facts and information.

The implicit and invisible cognitive biases that humans bring to every meeting every day hold us back. Scour your historical consciousness and you will find thoughts such as “I don’t like this guy, so all his opinions are garbage,” or “I am just going to follow what the boss says,” without really understanding everything. Human gut, instinct, impulse, habits are all driving us unconsciously, invisibly. Unfortunately you are not in control of the productivity wheel.

How you see the world, your behavior and decision-making is all unique and infinitely complex, but all patterned by high dimensional mathematics (a.k.a. AI). Synergizing your unique perspectives, ideation and problem solving with an artificial knowledge system like AI can produce incredible results in quality, accuracy and speed resulting in incredible “AI assisted” performance.

We see this confluence every day inside the company and with our customers by building group decisions automatically with each human voice being a part of the discussion to determine the best outcome. Consensus and collaboration without the human in the mix but in their unique voice through their unique prism and in synchronous cognitive sequence allows for easy ownership and accountability, at incredible speed.

And the final kicker that no one is seeing? What are you going to do when Claude or ChatGPT make a bad decision or make an error in judgment? Are you going to fire them? Write them up in Workday? Dock their pay? Good luck. It will never happen. Only humans can be held accountable.

And don’t even get me started on the Token Tsunamis happening right now. Left uncontrolled, companies are spending hundreds of millions of dollars beyond their annual budgets on AI coding and all for what? Where’s the ROI? And the prices for these models are only going higher. The small companies will be priced out very soon.

Luckily the solution is right in front of our noses: the mirror. At the end of the day, AI is only a tool. And a rather crappy tool that needs duct tape (RAG and vector databases), bailing wire (MCP and Frameworks and Agents), bubble gum (guardrails and governance), spittle (memory and state management) and Elmer’s wood glue (evaluation and synthetic testing) just to get a mostly correct answer to the question: “Who is the current US president?” on an offline local model.

All this scaffolding is in place today because the core LLM sucks at all the above. More tokens. More Token Typhoons.

AI today (at best) is Artificial Knowledge that can absolutely be leveraged by smart people and senior staff to produce high quality decision making and actioning. Do not get discouraged, but instead learn how to leverage it. Dive deep and differentiate yourself. Only through leveraging a good tool (like a skill saw) can you frame a house.

Interpersonal friction and endless alignment meetings are keeping your team from doing their best work. I make it my mission to unblock your team's highest potential, using our AI and turn ideas into decisions in minutes. My team and I harness AI for the betterment of humans to prevent workforce extinction. For now…

We've listed the best video conferencing software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

• Bad Apples has been given a global theatrical release date after premiering at TIFF
• A new trailer confirmed the movie would be in theaters from September 18 for viewers in the UK and US
• The dark comedy is the English-language debut for Jonatan Etzler

Bad Apples may have fallen under the radar after its TIFF debut, but the new dark comedy movie finally has a global release date.

The Saoirse Ronan led movie will be in UK and US theaters from September 18, and we've got a brand new trailer to celebrate the new movie's theatrical release thanks to Paramount Pictures.

Bad Apples follows Maria, a teacher who becomes frustrated when her class of ten year olds is disrupted by one unruly pupil, Danny (Eddie Waller). When Maria accidentally kidnaps Danny and holds him hostage in her house, it sets off an unexpected chain of events.

Take a look at the trailer below.

This is the kind of movie that reels me in instantly. We've got a seemingly ordinary main character in the form of a schoolteacher, who then goes on to make very questionable characters.

The plot of Bad Apples teases that Maria sees her class performance improve once Danny is "missing", which gives her a moral dilemma. Should she release him or keep him held in her home, where he's not disrupting the rest of her class?

Bad Apples is based on the novel De Oönskade (The Unwanted) by Swedish author Rasmus Andersson, which is described as a satirical thriller. So I'm expecting a lot of timely commentary in this movie, which will soon be widely available in theaters.

The movie currently has an 83% Rotten Tomatoes score, with critics praising the story, performances and soundtrack. So it's promising stuff already.

September 18 is a long wait, but thankfully we have plenty of recommendations to keep you entertained until then. If it's horror you're after, check out my round up of the best horror movies to stream in July.

Or, if you'd like more general Paramount movies, check out the best movies to watch on Paramount+.

• Palo Alto Networks CEO Nikesh Arora says 90% reduction in AI pricing is needed to drive widespread adoption, 20% reduction by next year
• OpenAI's latest GPT-5.6 Sol model is 54% more efficient across agentic coding tasks
• Shift to consumption-based models is making high costs even more pronounced

Palo Alto Networks CEO Nikesh Arora has warned that AI pricing is still too high for widespread enterprise adoption, arguing that it should be drastically cheaper.

Although model efficiency and performance have come a long way in recent years, he believes cost remains one of the biggest barriers for widespread adoption.

In a recent interview with CNBC, Arora laid out how he believes token prices need to fall by as much as 90%, however he acknowledges that the change is unlikely to happen immediately, instead welcoming a 20% price reduction over the next year or so.

Arora's comments come in response to OpenAI's latest announcement, revealing that its latest GPT-5.6 Sol model is now 54% mode token-efficient across agentic coding tasks – a major improvement.

Though Palo Alto Networks' CEO admitted that this in itself is a good start, it's not enough of an improvement to drastically change pricing for major enterprises at the moment.

"It's important to understand the demand continues to be infinite," he added, implying that cost remains prohibitive. The issue of cost is also becoming more of an issue as AI companies evolve their pricing strategies, with customers moving toward consumption-based models rather than flat monthly fees per user.

Importantly, Arora's vision for cheaper AI isn't unfounded. Virtually all previous technological revolutions have already followed a similar pattern. In the case of AI, compute would become cheaper, models would become more efficient and competition could keep prices low, all working in the favor of enterprise customers.

Decentralized, encrypted, and completely private messages. That's what Session promises.

You may never have heard of the app, but for its 1.7 million monthly users, Session is an important piece of privacy tech worth fighting to sustain.

Due to a severe funding shortfall, Session was expected to cease operations this week. However, after securing $200,000 to keep the project afloat, the team is now ready to look ahead.

Session co-founder Chris McCabe described the app as a "critical tool for fighting against an Orwellian future." But what is Session, and what does it offer that mainstream giants like Signal can't?

Launched in 2018, the project originally sustained its decentralized infrastructure through its own crypto tokens. According to McCabe, the Session Technology Foundation (STF) sold roughly $2.4 million worth of tokens toward the end of 2024 to keep the project going.

However, the model wasn't sustainable and, when a crisis in the crypto market drained Session's finances last year, the founders were forced to terminate the contracts of all 16 of their team members.

Unwilling to let the platform die, the founders scrambled for alternative lifelines. Following a high-profile emergency donation from Ethereum co-founder Vitalik Buterin — who gifted 128 ETH to support the app's mission — the foundation decided it was time to turn to its user base for long-term survival.

Without additional funding, Session's doors will close next month. Please read this appeal from Session co-founder Chris McCabe.https://t.co/nTJCocoDwIMarch 19, 2026

That grassroots fundraising campaign ultimately paid off.

To date, Session has received hundreds of community donations totaling $205,000. While that is still a long way off the project's ultimate $1 million sustainability goal, it is enough to keep the platform operational, albeit with far fewer resources than it once had.

"This funding allows us to retain three developers and one community operations person to maintain the app," McCabe explains.

"Send messages, not metadata" is Session's mantra. To back this up, the app uses a variety of advanced techniques to ensure users' messages remain private and secure.

From basics like end-to-end encryption to the use of a decentralized infrastructure and onion routing, Session goes significantly further than mainstream messaging apps.

"Session is completely trustless," McCabe told TechRadar. "You don't even have to trust us because the open-source code guarantees your security."

In the future, the app's design should also make it less vulnerable to increased surveillance and monitoring, such as the initiatives currently being proposed in Canada, the EU, and the UK.

In October 2024, the platform's creators shifted operations to Europe to help protect users' privacy. The move allowed the team to escape an Australian legal landscape that the founder described as "increasingly hostile" toward encrypted apps, relocating to the traditionally privacy-friendly jurisdiction of Switzerland.

Because the platform is open source, McCabe claimed the app can continue to operate even in countries where it has been banned from mainstream app stores, noting that users can easily bypass those restrictions by sideloading the app on Android devices.

And what happens if a repressive government demands data about an individual user? McCabe noted that the foundation simply couldn't comply, saying: "It's technically impossible for our company to copy or store that data."

Crucially, users do not need a phone number or email address to sign up for the service. That anonymity is especially beneficial for citizens in conflict zones or highly restrictive digital environments, where hostile governments can block SMS two-factor authentication messages to cut off access to mainstream messaging networks.

So, how exactly does Session stack up against mainstream giants like Signal?

Signal is widely recognized as the gold standard for privacy-preserving messaging apps. Like Session, it is fully open-source, collects minimal metadata, and uses advanced end-to-end encryption via the Signal Protocol.

While McCabe agrees that "Signal is great," he believes that being completely decentralized and eliminating the need for a phone number tips the scales in Session's favor.

However, not everyone in the cybersecurity community shares this view.

A prominent cryptanalyst who uses the pseudonym Soatok has publicly argued against switching from Signal to Session. A major point of criticism is Session’s decision to remove Perfect Forward Secrecy (PFS) — an important security feature that ensures that past chats remain private even if a user’s long-term identity keys are compromised in the future. Soatok also suggested that Session contained several other cryptographic flaws and implementation mistakes.

Bill Budington, Senior Staff Technologist at the Electronic Frontier Foundation (EFF), backed Soatok’s analysis, telling TechRadar that Session appears to be another aspiring Signal competitor that "comes up lacking on its security claims."

Mohammed Al-Maskati, the Digital Security Helpline Director at Access Now, shared this concern. He argued that sticking with a widely implemented standard like the Signal Protocol is generally a safer option because it significantly reduces the risk of making critical errors during development.

Responding to these critics, McCabe explained that while Session initially relied on the Signal Protocol, the team was forced to build its own encryption method because Signal's architecture wasn't designed to handle a decentralized node network.

However, he confirmed that PFS is officially slated to be reintegrated into the platform with the upcoming Session Protocol V2 upgrade.

Security experts have also pointed out that Signal remains a stronger choice for most users due to the sheer size of its network, which makes it far easier to use with friends and contacts.

However, McCabe rejects the idea that popularity equates to security, arguing that by that logic, WhatsApp would be considered a better privacy option than Signal. "Which it is not," he said.

As Session prepares to enter its next phase, the foundation is focused on achieving long-term financial sustainability.

To generate steady revenue, the platform plans to launch a premium subscription tier called Session Pro. Similar to Telegram Premium, Session Pro will offer power users advanced features — such as higher message character limits, expanded group sizes, more pinned chats, and animated profile pictures — backed by cryptographic proofs that preserve user anonymity during payment.

Most pressing, however, is finalizing Session's Protocol V2. Beyond reintroducing Perfect Forward Secrecy (PFS), it will also introduce post-quantum cryptography to secure users' chats against 'harvest now, decrypt later' threats.

While the wider InfoSec community continues to be divided on whether Session's custom architecture outperforms legacy rivals, the core team remains fiercely dedicated to the project.

"Many former team members are incredibly passionate and are taking pay cuts or donating their time to help get it back together," said McCabe.

And with mandatory age verification, lawful access to data, and encryption backdoors challenging the very notion of online privacy globally, it seems that messaging app like Session are going to become increasingly important.

As McCabe puts it: "People often don't realize they need this level of privacy until things go really badly and freedom of speech is restricted."

If you want to support the platform, Session continues to accept donations directly through its official website and mobile application.

Kimi is the AI assistant developed by Moonshot AI, a Beijing-based company that built its reputation on long-context processing long before that became standard. The platform runs on Kimi K2.6, an open-weight Mixture-of-Experts model released in April 2026, and it has made real inroads among developers looking for frontier-quality AI at a lower cost than OpenAI or Anthropic.

Two things set Kimi apart. Its API pricing sits around $0.55 per million input tokens against GPT-5.4's much higher rates, and its Agent Swarm architecture can coordinate up to 300 parallel sub-agents for large-scale automated workflows. The platform also covers the full office productivity stack with Slides, Docs, Sheets, and a website builder under one subscription.

We've been reviewing B2B software at Techradar Pro since 2012. You can also check out our other generative AI coverage, including our best AI tools roundup and our 2026 vibe coding buying guide.

Kimi is an AI chat assistant and agent platform from Moonshot AI. It launched in October 2023 with a focus on long-context processing, and its first version handled 128,000 tokens of context, which was exceptional at the time. The platform has since grown into a broader productivity suite covering deep research, document creation, coding, and multi-step agentic automation.

The underlying K2.6 model uses a 1-trillion-parameter Mixture-of-Experts architecture with around 32 billion parameters activated per token, keeping inference costs low while maintaining strong benchmark performance. It supports a 256K–262K token context window and processes text, images, code, and video natively.

Kimi targets developers, researchers, and teams running demanding workloads. Casual users get good value from the free tier, but the platform's design clearly aims at power users who need more than a chatbot.

Attribute

Notes

Underlying model(s)

Kimi K2.6 — 1T-parameter MoE, 32B active params, 256K–262K context, native multimodal

Best for

Deep research, long-document analysis, agentic coding, full-stack web generation

Distinguishing functions

Agent Swarm (300 sub-agents), Kimi Code, Deep Research, Claw Groups, WebBridge

UI features

Four response modes: Instant, Thinking, Agent, Agent Swarm; available on web, iOS, Android

Subscription costs

Adagio (Free), Moderato ($19/mo), Allegretto ($39/mo), Allegro ($99/mo), Vivace ($199/mo)

API pricing

Around $0.55/M input tokens and $2.65/M output tokens; OpenAI-compatible; context caching cuts input costs by up to 75%

• You need a capable coding assistant at a lower price point. Kimi Code, available from the Moderato tier upward, pairs K2.6 with terminal and VS Code integration and holds up well on multi-file refactors.
• Your work involves very long documents. The 262K-token context window lets you work through large contracts, codebases, or research papers in a single session without manually chunking content.
• You're building agent-driven automation. Agent Swarm on the Allegretto plan can coordinate up to 300 parallel sub-agents across 4,000 coordinated steps, a serious option for teams with research or data processing pipelines.

• You need live customer support. Kimi's support is self-serve through a help center and community channels, with no live chat or phone line.
• Data residency is non-negotiable. Moonshot AI is a Chinese company. Teams in regulated industries or with strict EU/US data sovereignty requirements should verify Kimi's data policies before committing.
• Agent Swarm is your main draw but you want entry-level pricing. Full agentic features only unlock at $39/month and above, while the free tier limits you to one concurrent task.

My first session was document-heavy. I dropped two long PDFs into the same conversation and asked Kimi to cross-reference specific sections. The results were accurate and well-organized, and the platform held up when I pushed with follow-up questions. That long-context handling is one of Kimi's strongest suits, accessible even on the free tier.

I also tested Kimi Code on a Python refactoring task. The output was clean, and when I asked it to explain architectural decisions, the reasoning held up. It's not quite at the same level as Claude Code for structured explanations, but the price difference makes the trade-off easy to accept for many workflows.

Agent Swarm on the Allegretto plan handled a competitive research task reasonably well, but the orchestration was not always transparent. A few sub-tasks returned incomplete results without flagging them as such, so I had to verify outputs manually. The platform has improved since K2.5, where tool call failures reportedly ran around 12%, but production users should still build in a review step.

Kimi's four response modes cover most use cases without requiring much configuration. Instant works for quick answers, Thinking for step-by-step reasoning, Agent for research and creation tasks, and Agent Swarm for parallel execution at scale.

Deep Research runs multi-step research workflows autonomously, pulling from web sources and producing structured reports. The output quality is good enough for first drafts, and the process shows which sources were used. Kimi Code supports multi-file editing and autonomous bug-fixing; Moonshot's documentation references coding sessions running continuously for over 13 hours on complex tasks.

The productivity suite covers Slides, Docs, Sheets, and Websites, giving Kimi a broader surface area than most chat platforms. The Websites tool generates full-stack sites from a single description, including frontend, backend, and database layers. These tools aren't the most polished in their respective categories, but having them integrated into one platform adds real convenience for teams with varied workflows.

The main Kimi interface is clean and web-first. The four response modes are clearly labeled, and mode-switching during a conversation is straightforward. A memory system carries context across sessions when you want it to. Mobile apps on iOS and Android work well for basic tasks, though agent workflows are better suited to desktop.

There's no structured onboarding tutorial, so new users will spend time with the help center. It covers everything from getting started through API usage across roughly 12 categories. Community channels on Discord and Reddit are active enough that you'll usually find someone who's hit the same issue.

Support is self-serve. The help center at kimi.com/help covers memberships, billing, API usage, and troubleshooting in detail. Discord and Reddit communities fill in the gaps for newer features.

What's missing is human support. There's no live chat, no prominently listed support email for general users, and no phone line. For a platform targeting developers running production workflows, that's a gap worth knowing about before you sign up. Enterprise arrangements may exist but aren't clearly documented publicly.

• Free (Adagio): Basic K2.6 access, limited agent usage, one concurrent task, no Agent Swarm, no Kimi Code quota.
• Moderato ($19/month): K2.6, Deep Research, Kimi Code, Slides, Websites.
• Allegretto ($39/month): Adds Agent Swarm and a higher professional data quota.
• Allegro ($99/month) and Vivace ($199/month): High-volume tiers; Vivace includes 30x Kimi Code usage and 240 swarm uses per month.

The free tier has no hard daily message limit as of mid-2026, which compares well to ChatGPT's free plan. For developers, API pricing is where Kimi's value is sharpest: around $0.55 per million input tokens and $2.65 per million output tokens, with an OpenAI-compatible endpoint. Context caching cuts input costs by up to 75% for applications sending repeated prompts, making it one of the more cost-efficient frontier options for teams building at scale.

• ChatGPT (OpenAI): The most widely used AI assistant, with a mature plugin ecosystem. Its $20/month Plus tier is roughly equivalent to Kimi's Moderato plan, but API costs are significantly higher for developers building at scale.
• Claude (Anthropic): Claude Pro at $20/month is a strong choice for document-heavy workflows and structured outputs. Claude Code is well-regarded for teams that prioritize explainability in AI-generated code.
• Gemini Advanced (Google): A natural fit for teams already in the Google Workspace ecosystem, with strong document workflow integration through Drive.

• Ran multi-document research tasks and stress-tested the context window with long technical files to evaluate accuracy across extended sessions.
• Tested the VS Code extension and CLI agent on Python and JavaScript tasks, including multi-file refactors and debugging workflows.
• Ran research and synthesis tasks using Agent and Agent Swarm modes on the Allegretto plan, checking output accuracy and how the platform handled incomplete results.

Kimi is available at kimi.com on a freemium model, with paid plans starting at $19/month. API access is available separately through Moonshot AI's open platform.

Mistral Vibe is one of the more interesting AI platforms to come out of Europe. The product, formerly called Le Chat, was rebranded by Paris-based Mistral AI in May 2026 to mark its shift from a conversational assistant to a full agentic platform covering chat, work automation, and cloud-based coding. It's a meaningful change, not just a naming exercise.

Two new operating modes, Work and Code, are the engine behind that ambition. Work Mode runs multi-step tasks across connected business tools. Code Mode handles remote coding sessions inside isolated sandboxes and delivers work through to a pull request. Together they put Vibe in more direct competition with ChatGPT and Claude than its Le Chat days suggested.

At TechRadar Pro, we've been reviewing business software since 2012 and our AI coverage has become some of our most-read work. That includes our AI tool roundup, our 2026 vibe coding buying guide, and deep dives on platforms like OpenClaw or Moltbook.

Mistral Vibe is a chat, work automation, and coding platform developed by Mistral AI. It's available via web browser and mobile apps on iOS and Android, and runs on Mistral's own model family, from the lightweight Small 3.1 to the flagship Large 3 and the reasoning-focused Magistral line.

The platform launched in February 2024 as Le Chat, originally a general-purpose assistant. Over the following two years, Mistral layered in web search, voice mode via its Voxtral audio model, a Canvas document editor, image generation through Black Forest Labs Flux Ultra, persistent memory, and project folders.

The May 2026 rebrand signals that Mistral sees Vibe as an enterprise-grade product. Work Mode and Code Mode are aimed at professionals who need an AI that can execute multi-step tasks across connected tools, not just hold a conversation.

Attribute

Notes

Underlying model(s)

Mistral Large 3, Medium 3.5, Small 3.1; Magistral for reasoning; Codestral and Devstral for code tasks.

Best for

Daily productivity, agentic work tasks, cloud coding, privacy-sensitive workflows.

Distinguishing functions

Work Mode, Code Mode, No Telemetry Mode, Canvas editor, Deep Research, voice mode, MCP support.

UI features

Web app and iOS/Android apps; unified interface for chat, canvas, and code views.

Subscription costs

Free; Pro at $14.99/month; Team at $24.99/user/month ($19.99 on annual billing); Enterprise (custom).

API pricing

Pay-per-token with no monthly minimums. Large 3 at $2/$6 per million tokens; Small 3.1 at $0.20/$0.60; Ministral 8B at $0.10/$0.10.

• You handle sensitive client or business data. No Telemetry Mode gives Pro subscribers a contractual guarantee that nothing they type is used to train Mistral's models. That level of assurance at $14.99 per month is unusual among major AI chat platforms.
• You want agentic features without enterprise pricing. Work Mode handles multi-step tasks across Gmail, Slack, Notion, and other connected tools. Code Mode manages full coding sessions through to a pull request. Both are available on the Pro plan.
• You're a developer watching API costs. Mistral Large 3 at $6 per million output tokens significantly undercuts GPT-5.4 ($15/M) and Claude Sonnet ($15/M) at the flagship tier, and the API bills only for tokens used.

• The free plan is your entry point for serious evaluation. At around 25 messages per day with no canvas and no remote coding access, the free tier doesn't give you a fair picture of what the platform can do.
• Your team is embedded in Microsoft 365 or Google Workspace. Vibe connects to business apps via MCP, but native integration with the major office suites is not as developed as what you'd get from Copilot or Gemini for Workspace.
• You're assessing the Team plan for a small group. The jump from Pro ($14.99) to Team ($24.99/user/month) is steep, and the main additions (admin controls and more storage) may not justify the cost for teams of fewer than ten people.

I tested Vibe across its Pro plan over several weeks, using it for research tasks, document drafting, and code work. The chat interface is clean and approachable. Anyone familiar with ChatGPT or Claude will navigate it without confusion. What caught me off guard was how efficiently Work Mode handled complex, multi-step research requests, pulling from connected sources and drafting a structured Canvas output in a single run.

Code Mode held up well for the tasks I threw at it. I ran a session to scaffold a simple API integration, and the agent handled writing, testing, and preparing a draft PR inside the sandboxed environment. I stepped in twice to give it additional direction, but that's consistent with what you'd expect from any AI coding agent at this stage.

The No Telemetry Mode stood out as a differentiator. Enabling it took seconds and gave me real confidence when working with business-related documents. That kind of data control is typically reserved for enterprise tiers at other major platforms, so finding it on a $14.99 plan is a real differentiator.

Vibe's feature set has grown considerably since the Le Chat launch in 2024. Alongside standard chat, the platform now covers web search, image generation, voice input, a Canvas document editor, Deep Research, memory, and project organization tools. Pro subscribers get access to Work Mode and Code Mode, which is where the real differentiation lies.

Work Mode is the headline addition from 2026. It turns Vibe into an execution agent that can read emails and calendars, draft documents in Canvas, run recurring scheduled tasks, and push outputs to Notion, SharePoint, or Slack. Every step is visible in the interface. The platform asks for explicit approval before any action that modifies data or sends a message.

Code Mode targets developers specifically. It connects to GitHub, GitLab, Jira, and Linear, running sessions in an isolated sandbox. A /teleport command lets you move a session between the Vibe web app and a local terminal without losing context. Parallel sessions are supported on Pro and above, which is useful for running background jobs while staying in another workflow.

Memory, currently in beta, lets Vibe store your preferences and recurring context across conversations. You can view, edit, or delete these entries at any time, and turning the feature off is a single toggle in confidentiality settings. Image generation via Black Forest Labs Flux Ultra is solid for a chat platform.

The main gap in the feature set is transparency around usage limits. The free tier runs at lower quotas than Pro across searches, image generations, and messages, but Mistral doesn't publish the exact numbers on its pricing page. That makes the free-to-Pro comparison harder to assess than it should be.

Overall, Vibe covers more professional use cases than most platforms at this price point. The area where competitors like Copilot and Gemini maintain a clear lead is native productivity suite integration, which Mistral has not yet matched.

The layout is straightforward: a left sidebar for conversation history and project folders, a central chat window, and model or tool selectors accessible from the input bar. Switching between chat, Work Mode, and Code Mode happens within the same interface rather than routing you to a separate product URL.

The mobile apps mirror the web experience closely. Search, canvas, image generation, and voice input all carry over, which isn't a given with AI chat platforms on mobile. Onboarding is minimal, which suits experienced AI users but may leave newcomers without much guidance on how to get started with the more complex agentic features.

Free and Pro users access support through the help center widget on the Mistral site. Response times are not published, and there's no live chat or phone support at these tiers. The help documentation covers the most common issues in reasonable depth, but for billing questions or edge-case technical problems, you're relying on ticket-based email support.

Enterprise customers get a dedicated support workflow with priority routing. Requests go through the same widget but are flagged and handled separately based on account type. For teams in finance, healthcare, or other regulated sectors where response time matters, that distinction is a real consideration when deciding between Team and Enterprise.

• Free plan: Around 25 messages per day with limited web searches, reduced image generation, no canvas creation, and no remote coding.
• Pro at $14.99/month: No Telemetry Mode, full canvas access, Work and Code Mode, 5x more web searches, more image generations, and pay-as-you-go Vibe coding beyond included limits.
• Team at $24.99/user/month (or $19.99 on annual billing): Shared workspaces, admin controls, domain verification, data export, and higher storage limits.

The free plan gives you a taste of chat quality and basic search, but the restrictions mean you won't get a representative experience. Pro at $14.99 is, by most comparisons, the cheapest premium AI chat subscription from any major provider. The No Telemetry Mode and agentic modes are hard to find elsewhere at this cost.

The Team plan's value depends on your use case. For small teams, the mainly administrative additions over Pro may not justify the per-user cost. Enterprise pricing is negotiated directly with Mistral and covers SAML SSO, on-premise deployment, custom model training, and dedicated support. The API runs on a fully separate billing track with no monthly minimums, making it accessible for developers at any scale.

• ChatGPT (OpenAI): The most established option for general productivity, with deeper Microsoft 365 integration and a broader ecosystem of plugins. It costs more at comparable tiers but suits teams already embedded in the Microsoft environment.
• Claude (Anthropic): A strong choice for long-document analysis and nuanced writing. Pricing is competitive and reasoning quality is high, though data privacy controls at the consumer tier are less explicit than Vibe's No Telemetry Mode.
• Gemini for Google Workspace: The best fit for teams already using Google's suite. Native Calendar, Docs, and Gmail integration outpaces what Mistral currently offers for Google-centric workflows.

• Ran prompts across standard chat, Web Search, and Deep Research modes to evaluate response accuracy, source quality, and multi-step research handling across a range of topics and document types.
• Tested agentic task execution through Work Mode, including document drafting via Canvas and recurring task scheduling, and ran coding sessions through Code Mode against a GitHub-connected project to assess end-to-end agent performance.
• Verified plan details and API rates against Mistral's official pricing page, cross-referenced them with third-party pricing analyses where official documentation was vague, and assessed support options through the Mistral Help Center.

I tested Mistral Vibe on its Pro plan over several weeks using a mix of daily productivity tasks and structured feature evaluations. Pricing data was verified against the official mistral.ai pricing page, with third-party sources used for cross-reference on API rates and plan limits where Mistral's documentation was unclear. Support quality was assessed through available help center documentation and publicly reported user experiences.