SUCCESS

Meta’s newest AI model Muse Spark 1.3, unveiled yesterday, is faster and more performant on third-party benchmarks than its predecessor — with a caveat. “Muse Spark 1.3 is rolling out today with frontier performance almost too cheap to meter,” Meta co-founder and CEO Mark Zuckerberg wrote on X, calling it Meta’s “biggest jump” yet in coding and agentic work.There is substance behind both parts of that claim. Muse Spark 1.3 makes significant gains over last month’s 1.2 release, particularly on long-running agent tasks. The version developers can access now is also one of the strongest price-performance offerings near the top of independent model rankings.Meta’s strongest Muse Spark 1.3 benchmark results come from its max reasoning configuration. Meta says that version is still completing additional safety testing and will arrive “shortly”; the third-party benchmarking firm Artificial Analysis says it evaluated max in a limited partner preview, and currently lists no API provider at all for the configuration. The version broadly rolling out this week through its Muse Code harness and the Meta Model API uses Meta’s previously available reasoning settings, including xhigh.That makes the more relevant enterprise question not whether Muse Spark 1.3 can reach frontier territory, but how close the model companies can actually deploy today gets — and at what real cost.The shipping model is very good, but not the benchmark leaderMeta does disclose results for both configurations in its underlying evaluation report, so this is not a case of the company hiding the deployable model. But its launch materials prominently showcase the max variant, and some of the largest scores belong to that configuration.For example, Meta reports GDPval-AA v2 scores of 1,754 Elo for max versus 1,709 for xhigh, OSWorld 2.0 scores of 66.9 versus 57.2, and JobBench scores of 64.9 versus 61.2. On some tests the distinction is negligible or reversed: DeepSearchQA is tied at 89.4, while xhigh scores 89.2 on Terminal-Bench 2.1 versus max at 88.8.Artificial Analysis scores Muse Spark 1.3 max at 62 on its Intelligence Index and the shipping xhigh version at 61. The latter ties GPT-5.6 Sol max, Grok 4.6 high and Claude Opus 5 high. But Anthropic still occupies the top of the leaderboard: Claude Fable 5.1 reaches 66 at max and 65 at xhigh, while Claude Opus 5 reaches 63 at max and xhigh.In other words, Muse Spark 1.3 xhigh is legitimately in the frontier cluster, but it is not the model currently setting the frontier.That is still a substantial change from Muse Spark 1.2. VentureBeat’s coverage of last month’s launch found Meta fielding a credible coding challenger that nevertheless generally trailed Anthropic’s best model. Muse Spark 1.2 scored 82.9% on Terminal-Bench 2.1 versus Opus 5’s 86.7%, and also finished behind Opus on the other main coding comparisons Meta presented.With 1.3, Meta is no longer merely showing up in that contest. On several coding and agentic evaluations, it is trading wins with OpenAI and Anthropic.Meta says the underlying model has also become easier to operate. Muse Spark 1.3 is trained to maintain multiple workflows in a long thread, gather context with tools, detect gaps in its own plans, ask users for clarification when necessary and confirm before consequential actions. In Meta engineers’ internal comparisons, it used roughly 20% fewer tool calls and 25% fewer tokens than 1.2 during coding work.For enterprises paying for thousands or millions of agent loops, those behavioral improvements could matter more than another leaderboard point.‘Almost too cheap to meter’ does not mean Meta cut its pricesMuse Spark 1.3 did not receive an API price cut. Meta kept Standard pricing exactly where it was for Muse Spark 1.2: $1.25 per million input tokens, $4.25 per million output tokens and $0.15 per million cached input tokens.ModelInput ($/1M)Output ($/1M)Total ($/1M)SourceMuse Spark 1.2 / 1.3 Contributor$0.10$0.20$0.30MetaMiMo-V2.5 Flash$0.10$0.30$0.40XiaomiDeepSeek-V4-Flash — off-peak$0.22$0.66$0.88DeepSeekGPT-5.6 Luna$0.20$1.20$1.40OpenAIMiniMax-M3$0.30$1.20$1.50MiniMaxLongCat-2.0 — limited-time promo$0.30$1.20$1.50LongCatDeepSeek-V4-Flash — peak hours$0.44$1.32$1.76DeepSeekMiMo-V2.5$0.40$2.00$2.40XiaomiDeepSeek-V4-Pro — off-peak$0.66$1.98$2.64DeepSeekLongCat-2.0 — standard$0.75$2.95$3.70LongCatMiMo-V2.5 Pro (≤256K)$1.00$3.00$4.00XiaomiGemini 3.7 Flash — through Dec. 31, 2026$0.75$3.75$4.50GoogleGemini 3.8 Flash — through Dec. 31, 2026$0.75$3.75$4.50GoogleDeepSeek-V4-Pro — peak hours$1.32$3.96$5.28DeepSeekMuse Spark 1.1 / 1.2 / 1.3$1.25$4.25$5.50MetaGLM-5.3$1.40$4.40$5.80Z.AIGrok 4.6 — 256K)$2.00$6.00$8.00XiaomiQwen3.8-Max$2.00$6.00$8.00QwenCloudGemini 3.7 Flash — starting Jan. 1, 2027$1.50$7.50$9.00GoogleGemini 3.8 Flash — starting Jan. 1, 2027$1.50$7.50$9.00GoogleGPT-5.6 Terra$2.00$12.00$14.00OpenAIGrok 4.6 — ≥200K prompt tokens$4.00$12.00$16.00xAIGPT-5.4$2.50$15.00$17.50OpenAIKimi K3$3.00$15.00$18.00Moonshot AIClaude Opus 5$5.00$25.00$30.00AnthropicSakana Fugu Ultra (≤272K)$5.00$30.00$35.00Sakana AIGPT-5.6 Sol — Standard mode$5.00$30.00$35.00OpenAIClaude Fable 5 / Claude Mythos 5$10.00$50.00$60.00AnthropicClaude Fable 5.1 / Claude Mythos 5.1$10.00$50.00$60.00AnthropicGPT-5.6 Sol — Fast mode$10.00$60.00$70.00OpenAIThat makes Zuckerberg’s “almost too cheap to meter” line less a statement about lower token prices than about what Meta believes developers can accomplish with those tokens.Artificial Analysis offers evidence for that argument, but also a complication. It measures Muse Spark 1.3 xhigh at 235.2 output tokens per second and estimates a cost of $0.55 per Intelligence Index task. At 61 on the Intelligence Index, that gives it the lowest cost per task of any currently measured model at that intelligence level. Muse Spark 1.2 cost only $0.40 per Artificial Analysis task, while scoring 57. Despite unchanged per-token pricing, the independent benchmark’s cost of completing an average task therefore increased generation-over-generation. Artificial Analysis attributes the increase primarily to heavier input-token consumption on agentic evaluations. That does not directly contradict Meta’s claim of 25% lower token use: Meta is describing comparisons in its own coding workflows, while Artificial Analysis is measuring a broader suite of reasoning and agentic tasks. But it illustrates why “cheap” becomes slippery once models operate as agents. Token rates, reasoning effort, number of turns, tool calls and retries all contribute to the actual cost of finishing work.Meta also retains its unusually cheap Contributor tier — $0.10 per million input tokens and $0.20 per million output tokens — in exchange for permission to use prompts and completions for training. As VentureBeat noted with Muse Spark 1.2, that may be attractive for prototyping but creates a materially different data-governance calculation for enterprises working with proprietary code or sensitive internal information.Wang’s ‘Gemini who?’ lands on an unusually close comparisonMeta chief AI officer Alexandr Wang was considerably less qualified in celebrating the release.After Artificial Analysis posted its Muse Spark results, Wang reposted them on X, adding: “i really hate to say it, but… gemini who? 😱💨”The shade was particularly pointed because Google released Gemini 3.8 Flash on the same day, pitching it at almost exactly the same class of workload: long-horizon software engineering, autonomous agents and multi-step professional reasoning. Google calls 3.8 its best reasoning and coding Flash model yet and says it is the company’s third Flash release in six weeks. Independent numbers give Wang something to work with, though hardly a knockout.Artificial Analysis gives Muse Spark 1.3 xhigh a 61 Intelligence Index score at $0.55 per task, compared with 59 and $0.58 for Gemini 3.8 Flash at high reasoning. Meta therefore edges Google on both intelligence and task cost at those particular settings. Google wins decisively on throughput. Artificial Analysis measures Gemini 3.8 Flash high at about 305 output tokens per second, versus 235 for Muse Spark — roughly 30% faster. Gemini also has the lower raw API sticker price for now: Google is charging an introductory $0.75 per million input tokens and $3.75 per million output tokens, compared with Meta’s $1.25 and $4.25.That promotional Google pricing expires December 31, after which it rises to $1.50 per million input tokens and $7.50 per million output tokens. The result is a useful snapshot of how tight frontier-model economics have become. Meta currently wins this independent comparison by two Intelligence Index points and three cents per benchmark task; Google offers substantially higher output throughput and cheaper raw tokens during its launch promotion.Wang’s “gemini who?” is fun executive trash talk. For an enterprise architect, the answer is closer to: Gemini is the faster option; Muse is currently the slightly stronger high-effort agent by this independent measure.Meta’s evolving open weights stanceThe more consequential issue for some developers may have little to do with today’s benchmark race.When Meta launched Muse Code and Muse Spark 1.2 in August, VentureBeat noted how dramatically the company had moved away from the open-weight strategy that made Llama ubiquitous. Muse Code and Spark 1.2 were proprietary, API-served products — a striking posture for the company that had spent years arguing that open AI was the path forward.Five days later, Meta changed course again.On August 10, it released the 30-billion-parameter Muse Glimmer under an Apache 2.0 license. Zuckerberg also said: “In the coming weeks, we are also going to open the weights for Muse Spark 1.2.” Reuters separately reported Meta’s plan to release the Spark 1.2 weights. Now, Meta has instead shipped Muse Spark 1.3 as another proprietary model.That does not yet amount to a broken promise — “coming weeks” can reasonably describe a period longer than three weeks. But today’s announcement makes the roadmap less clear rather than more.Meta’s new post no longer says Muse Spark 1.2. It says its roadmap includes “the Muse Spark open weights release”, without identifying a version, release date, model size or license. Zuckerberg likewise said on X that “Muse Spark open weights releases” are coming soon. For teams that standardized on Llama because downloadable weights meant self-hosting, customization and control over inference economics, that ambiguity may matter more than whether Spark gained another point on a composite benchmark.Muse Spark 1.3 shows that Meta can now iterate proprietary frontier models at extraordinary speed. The shipping xhigh configuration is fast, competitively priced and much closer to the top of independent rankings than its predecessors. The max preview shows Meta can push the family a little further when allowed to spend more reasoning compute.The next test is different: whether Meta can convert that pace into a roadmap enterprises can actually plan around — including making its best capabilities broadly deployable and delivering the open-weight Spark model it has already said is coming.

Roku (ROKU) has spent years trying to own the software layer of television. Now it is getting more aggressive about owning the television itself.

Roku has launched its first OLED TVs, pushing into a premium display category long dominated by Samsung, LG, and Sony.

The new Roku Pro Series OLED starts at $999.99 for a 55-inch model, while the 65-inch version costs $1,199.99. Roku is also preparing a brighter Pro Series LX OLED for October, priced at $1,299.99 for 55 inches and $1,599.99 for 65 inches.

Roku is pushing OLED image quality farther down the market, but still with features that are more typical of more costly TVs: 120Hz refresh rates, four HDMI 2.1 connections, Dolby Vision, HDR10+, FreeSync Premium, and variable refresh rate gaming compatibility.

But Roku’s business model makes the launch more fascinating than a simple hardware pricing battle.

More than 100 million streaming households use Roku OS devices. Licensing partners, not Roku, make the majority of its TV sales.

This implies that Roku doesn’t necessarily need to earn giant profits by selling an OLED TV.

Another Roku OS screen is required.

Roku is attacking OLED at a difficult moment for TV makers

The timing is unusual, since global television demand is hardly booming.

TrendForce anticipates global TV shipments will fall 0.8% in 2026 to roughly 194.65 million units, while shipments in the first half would rise 1.3% to 93.74 million. The sector is under pressure from growing memory prices, softening demand, and shrinking margins for smaller producers.

Memory’s share of a television’s bill of materials is expected to jump from roughly 2.5% to 3% historically to 6% to 7%.

Related: A $650 smartphone takes aim at Apple and Samsung’s surging repair costs

That is a surprisingly big shift for a category where manufacturers already compete fiercely on price.

And the market is becoming more divided.

Samsung delivered 17.6 million TVs in the first half of 2026, up 6.3%. TCL delivered 15.08 million, up 7.1%; Hisense delivered 14.23 million, up 3%. LG delivered 11.3 million units, up 3.9 percent.

Against those giants, Roku’s manufacturing scale is comparatively modest. So why enter OLED now? Because Roku may be pursuing a different strategy.

The TV itself may be Roku’s customer-acquisition cost

Roku said something telling earlier this year.

Devices revenue was $118 million in the first quarter, down 16% year over year, with a negative 16.3% gross margin. The company attributed the drop to lower player sales and promotional pricing.

More Tech:

Anthropic-powered AI model sends shocking message to employee

Rocket Lab clears 1st hurdle in its biggest satellite deal

Apple’s $54 billion iPhone machine may be about to break its biggest ritual

That is not how most investors want a hardware business to look. But Roku’s TVs and streaming boxes serve another purpose. They get Roku OS into homes.

Once Roku has the operating system on the device, it can monetize the device over its lifetime through advertising, streaming subscriptions, content distribution, and other platform services.

That alters the economics of selling a TV.

Samsung, Sony, and LG usually require their premium TVs to be substantial hardware goods.

The transaction brings a long-term platform user, which might possibly help Roku overcome poor hardware economics.

Its first OLED TV is therefore more of a 55-inch screen customer acquisition device. That’s likely why Roku can afford to be so aggressive with the launch pricing.

Roku just found a new way to fight Samsung and LG.Bloomberg / Getty Images

OLED gives Roku access to a more valuable type of customer

OLED is still a top-end technology.

Traditional LCD TVs use LCD pixels that need a backlight, whereas OLED pixels make their own light, enabling deep blacks, excellent contrast, and incredibly thin displays. The category has always demanded costs much above mainstream market LCD TVs.

And although the overall TV industry is in decline, OLED is one of its growing sectors.

Omdia estimates that large-area OLED display shipments will grow 18.8% to 38.8 million units in 2026, while total large-area display shipments will fall 2.3%.

Not all of that increase is coming from TVs; monitors and laptops are contributing heavily. Still, it shows a larger transition to OLED even as weaker display technologies see reduced demand.

The move into OLED also affects the sort of home Roku can target.

Its brand has long been synonymous with cheap streaming sticks and budget televisions.

A Roku television priced between $1,000 and $1,600 brings the business into living rooms where customers are ready to pay substantially more for entertainment devices.

Those homes may also be very lucrative advertising and subscription clients. And that’s where the strategic payout might be higher than the margin on the TV.

Roku’s new television makes the software battle harder to ignore

For the most part, the television business was a fight for image quality.

The fight over what occurs when the TV goes on is becoming more of a war.

As the market moves away from hardware requirements alone, TrendForce specifically expects smart-TV platforms, advertising services, and content ecosystems to become more significant to manufacturers’ competitiveness.

That observation fits Roku unusually well. Roku started with the platform. The hardware came later.

The traditional TV makers have largely gone the other way: They made displays and then built operating systems, advertising platforms, and content ecosystems around them.

That makes Roku’s OLED growth strategically crucial.

Every premium Roku television sold has the potential to usurp the default gateway to a household’s streaming habit from a Samsung Tizen, LG webOS, or Google TV interface.

And the living room screen is the prime piece of real estate.

Consumers may hold on to a TV for several years. During that time, the operating system may show advertisements frequently, propose content, and enable subscriptions.

The $999 purchase happens once, yet the platform relationship can last much longer.

Amazon exclusivity adds another twist

Amazon is the only place to get Roku’s Pro Series OLED for 2026.

That provides Roku quick access to one of the biggest consumer electronics stores in the U.S. without the need to create a similar physical retail presence.

The basic Pro OLED now comes in 55- and 65-inch sizes. The more expensive LX, arriving in October, has around double the brightness, a 144 Hz variable refresh rate, and a polarizer meant to cut down on reflections, according to product specs provided by Roku.

Roku’s official product page confirms the prices and OLED specifications. This is great for consumers. Buyers now have another option besides Samsung, LG, and Sony as OLED prices drop. For Roku investors, calculations differ. What matters isn’t whether Roku can become America’s largest OLED maker. Perhaps it’s unnecessary.

Roku OS is already running in more than 100 million streaming homes. The trick is keeping those families interested while adding new ones and extending the platform into more lucrative portions of the television industry.

An inexpensive OLED provides Roku another option to accomplish it. It could also explain the unconventional economics of the launch.

Roku isn’t simply trying to sell you a cheaper premium television. It may be willing to make the television cheaper because it really wants to own everything you do after you switch it on.

Related: Apple’s new iPhone strategy comes with a $2,500 question

Gold rose as much as 3% on Thursday, while silver surged nearly 4%.

Though Trump has refused to say where the tariff-free beef is coming from, the specifics of global beef trade point directly to Brazil as the primary source.

Michael Burry made his name betting against a market everyone else trusted. This time, the market pushed back harder than usual, and the damage showed up across nearly every corner of his bearish portfolio.

August turned into one of the roughest months yet for his bearish AI positions, even as he kept adding to them rather than backing away. The stretch offers a real test of how long conviction can hold up against a genuinely strong earnings season.

Michael Burry’s bets against Palantir Nvidia and Micron backfired in August

Most of the stocks Burry was bearish on advanced during August, led by sharp gains in Palantir, Nvidia, and Micron. The month highlighted a real risk for AI skeptics: valuation concerns can look extreme for a long time before momentum and earnings growth actually break. August gave little evidence that a break is imminent.

Palantir was by far the biggest problem. The stock surged 51.4% in August, even as Burry maintained out-of-the-money put options with a $100 strike expiring December 2026 and a $50 strike expiring June 2027, according to Yahoo Finance.

More AI:

Nvidia just made a move Wall Street wasn’t ready for

Microsoft just took sides in AI policy fight

OpenAI just disclosed something genuinely alarming

Nvidia and Micron added their own pressure. Nvidia rose roughly 8% in August while Burry held bearish put positions alongside higher-strike December calls meant to offer some protection around earnings. Micron gained 13.3% even after Burry increased his short exposure, arguing the memory maker’s rally reflected speculation more than fundamentals, according to Stocktwits.

Other names in Burry’s bearish basket told a similar story. Oracle climbed 16.2%, CoreWeave gained 17.4%, Tesla advanced 12.1%, and Nebius rose 9.9% over the month, leaving Applied Materials and Caterpillar as the only two of his tracked short targets that actually declined, down 9% and 1.8% respectively, Stocktwits noted.

Palantir was the biggest problem in Burry’s portfolio

Palantir’s second-quarter results, reported August 3, gave the stock’s rally real fundamental backing. Revenue reached $1.935 billion, up 93% year over year, with U.S. commercial revenue growing 149% to $764 million and adjusted free cash flow of $1.22 billion, TheStreet reported.

Burry has not backed off his long-term view despite those numbers. He has said Palantir is worth under $1 per share on a long-term basis, a stark contrast to a stock trading near $175 with a market capitalization around $420 billion and a trailing price-to-earnings ratio above 137, according to Stock Analysis.

Burry’s argument is not that Palantir’s business is weak. He has acknowledged the company’s Rule of 40 score, a metric combining revenue growth and profit margin, reached 155% last quarter, far above the benchmark for AI infrastructure companies. His case instead is that the current price already assumes near-perfect execution indefinitely, leaving little room for anything to go wrong, as TheStreet reported.

In early August, Burry widened the comparison further. He likened the current AI buildout to 2005, one year before the housing market began to crack. He also disclosed fresh shorts against Oracle and Nebius around that same period, broadening his bet beyond the three most talked-about names.

Most of the stocks Burry was bearish on advanced during August, led by sharp gains in Palantir, Nvidia, and Micron.Michael/Getty Images

Why Michael Burry still isn’t backing down on his shorts

Burry’s positioning has grown rather than shrunk as the trades have moved against him. His short positions in the iShares Semiconductor ETF, Micron, Nvidia, Caterpillar, Palantir, Tesla, and Applied Materials remained largely intact up until early August. He has said publicly that all of those positions stayed profitable except his bet against Nvidia, according to TheStreet.

Micron’s own business results complicate Burry’s timing argument. Chief business officer Sumit Sadana said on the company’s fiscal third-quarter earnings call that customer demand for memory chips remains “well above our ability to supply” across nearly every product category through 2028, a comment that cuts directly against the idea that current demand reflects speculation rather than end-customer need, according to Insider Monkey.

Burry has also pointed to a more technical concern about how AI infrastructure spending gets accounted for. He has argued that hyperscalers may be extending the useful life of Nvidia chips in ways that understate depreciation and inflate reported earnings, estimating the resulting shortfall could reach roughly $176 billion between 2026 and 2028.

Burry’s call options on Nvidia, bought as a hedge rather than a bullish trade, at least partly cushion the impact from Nvidia’s own August strength. “I am not playing for gains here,” Burry wrote of that trade as the stock jumped roughly 8% following its August earnings report.

What investors should watch next after Burry’s August losses

August illustrates the gap between identifying valuation risk and successfully timing it. Burry may ultimately be right that parts of the AI trade are overpriced, but Palantir, Nvidia, and Micron all continue to benefit from genuinely strong earnings and sentiment momentum that has outpaced his bearish thesis so far.

Burry’s short positions are not a signal by themselves. The signal is the earnings data. Right now that data says AI demand is real and growing. If a major hyperscaler comes out next quarter with weak guidance or a spending cut, that changes everything. Nothing like that has shown up yet.

Burry has been here before. He was early on the housing trade too, and early felt a lot like wrong for a long time. August is one data point. The numbers that matter are the ones that come out of the next round of earnings calls. Until the demand story cracks there, the market is not listening.

Related: Michael Burry just sent a fresh signal to stock market investors

Bad Bunny’s “EOO” returns to its all-time high of No. 2 on Billboard’s Latin Streaming Songs chart as Karol G’s latest smash blocks him from a new champion.

The company’s AI-powered coding assistant is driving major sales momentum.

Nvidia (NVDA) has spent the AI boom selling companies the computing power needed to generate things that didn’t exist before.

Now it wants your graphics card to do something similar while you’re playing a video game.

Nvidia’s new AI-powered graphics innovation DLSS 5 is officially rolling out on Sept. 3, starting with NBA 2K27. Nvidia claims it is the company’s largest computer-graphics advance since real-time ray tracing was introduced in 2018.

Nvidia calls it 3D-Guided Neural Rendering. The neural model analyzes scene objects to add more realistic lighting, materials, shadows, and other visual details.

The result sounds like an obvious technological upgrade. It also raises an unusual question for gamers and, eventually, Nvidia investors: At what point does a graphics card stop rendering a video game and start generating one?

Nvidia has dramatically reduced DLSS 5’s hardware requirements

But perhaps the most astounding figure in Nvidia’s release has nothing to do with picture quality.

When Nvidia demonstrated DLSS 5 in March, the technology required two flagship GeForce RTX 5090 graphics cards.

Nvidia believes DLSS 5 performance is 5x better and can run on a single RTX 50-series GPU in six months.

That includes the RTX 5060.

Nvidia states that just about every RTX 50-series GPU can run NBA 2K27 at 1080p with ray tracing and the Ultra option, utilizing DLSS 5 and other DLSS technologies.

Related: Bank of America doubles down on Nvidia stock

At the extreme end, Nvidia says an RTX 5090 can reach up to 370 frames per second at 4K with Ultra settings, ray tracing, and DLSS.

At 1440p, Nvidia lists the maximum performance as:

RTX 5090: 590 fps

RTX 5080: 410 fps

RTX 5070 Ti: 350 fps

RTX 5070: 260 fps

Those figures need some context.

DLSS is not only making the GPU do the work of rendering hundreds of full frames per second in the traditional sense. With Multi Frame Generation, AI creates extra frames between the ones that Nvidia normally renders. Current DLSS technology can create up to five more frames for every displayed frame.

DLSS 5 crosses another line

That’s not contentious about DLSS 5 though: frame generation is.

The major difference is in what occurs in those frames.

Nvidia claims its neural model understands scene objects and semantics and can produce more realistic lighting, material reactions and subsurface scattering, while consuming information from the underlying 3D environment. Developers have controls over structure and tone and semantic masks to govern how aggressively the algorithm modifies the picture.

Nvidia claims DLSS 5 in NBA 2K27 can adjust how light goes through a player’s ears, enhance the look of skin and facial hair, and produce more nuanced shadows on jerseys and basketballs.

Those may sound like tiny details. But they represent a significant philosophical change in how video-game graphics can be created. Traditionally, developers create an asset and the GPU renders it.

Enter DLSS 5, a new AI model that can look at that item and create more visual information around it in real time.

That difference is part of why there was a pushback to Nvidia announcing DLSS 5 back in March.

More Nvidia:

Nvidia just made a move Wall Street wasn’t ready for

Nvidia just locked down deal that changes AI race

Nvidia stock is doing something it hasn’t done in years

Some of the earlier demos by Nvidia were derided for modifying characters too much, notably the look of Grace Ashcroft from Resident Evil Requiem, The Verge reported. The article likened the impact to running the game through a generative-AI filter in real time.

NBA 2K27’s AI is more focused on texturing, lighting, and shadows, according to The Verge.

Developers may also instruct DLSS 5 to leave certain sections, like faces, alone.

Nvidia’s newest AI feature comes with one ugly catchMATT RAMEY / Getty Images

Nvidia has another reason to make AI graphics indispensable

There is a commercial narrative behind the technical argument.

Nvidia’s Gaming unit recorded record sales of $16 billion in fiscal 2026, a 41% increase year-over-year.

This is a huge gaming company.

It is also modest in relation to what AI infrastructure has turned into for Nvidia.

The company just reported $96.2 billion of total revenue in a single quarter, up 106% year over year. Data Center alone generated $89 billion, up 117%.

It is an astounding comparison.

Nvidia’s last quarter of Data Center revenue was almost 5 times the total Gaming revenue for the entire prior fiscal year.

That doesn’t imply gaming is dead.

Instead, Nvidia is increasingly bringing the same basic economic proposition that transformed its data center business: more AI requires more Nvidia computing power, back to consumers.

DLSS 5 is one such case.

The technology is officially launching on RTX 50-series desktop and laptop GPUs and on Nvidia’s GeForce Now service. Nvidia said the next model upgrades should increase the performance further.

So, for a gamer with older hardware, some of the latest visual features are not just unlocked with a software update.

They can become another reason to upgrade the GPU.

Nvidia could change what a graphics upgrade actually means

For decades, PC gamers have been upgrading their graphics cards for a rather obvious reason.

A faster GPU would be able to produce more polygons, higher quality textures, sharper shadows and ultimately more realistic ray-traced lighting.

AI alters that equation.

If the neural networks can intelligently rebuild, augment or synthesize sections of the final product, the graphics card doesn’t have to brute-force every pixel and every frame in the conventional sense.

This might make the capabilities of AI more and more essential when buyers compare GPUs.

And Nvidia has been building just that edge into RTX.

Related: OpenAI’s agents breached Hugging Face. Nvidia wants it.

The RTX 50 series is based on the fifth generation of Nvidia’s Tensor Cores, a dedicated AI accelerator silicon. DLSS provides Nvidia with a consumer-facing application that can take those AI cores and convert them into something players will really see.

The strategic value could be more than one generation of graphics cards.

If game creators are able to start developing games with neural rendering at the core, rather than as an optional addition, AI technology might become as crucial to gaming GPUs as conventional rendering performance.

One game will now test Nvidia’s biggest graphics claim in years

There is a catch. At launch, the showcase is remarkably narrow. NBA 2K27 is the only confirmed game shipping with DLSS 5 in Sept. 3.

The games announced at Nvidia’s March event included Resident Evil Requiem, EA Sports FC, Starfield, and Hogwarts Legacy; however, The Verge reports that Nvidia did not provide an updated timeframe for DLSS 5 support in those titles.

That is, investors and gamers should not mistake the promise of the technology with the possibility for mass adoption.

Nvidia has shown that DLSS 5 can operate on much less hardware than it could six months ago.

Developers or players have yet to show that they want AI to change depicted scenes in broad strokes.

That’s why NBA 2K27 is more significant than it looks.

The discussion isn’t actually about whether artificial intelligence can make virtual bacon shinier or basketball players’ skin more lifelike.

It’s about where computer graphics are heading, according to Nvidia.

For decades, the graphics industry competed over which GPU could render reality most convincingly.

Nvidia is beginning to suggest that the next winner may be the GPU that can intelligently generate the missing pieces of reality instead.

Gamers can decide on Sept. 3 if that’s the future or if Nvidia’s AI has gone too far.

Related: Nvidia stock flashes unusual signal for investors 

When development companies put their funds behind a project, there is always the risk that things will not go the way they intended.

Two resorts in Miami Beach filed for Chapter 11 bankruptcy within a few weeks of each other in March 2026, while at the end of last year, the company behind well-known New York City hotels such as The Tuscany and Hotel 27 shut down operations in a situation that left guests from different parts of the world struggling to find last-minute accommodations at Manhattan rates.

The latest development company to file for Chapter 11 bankruptcy in the U.S. Bankruptcy Court for the Southern District of New York is the beleaguered Park Place Partners Development LLC.

Park Place Partners Development files for Chapter 11 bankruptcy

The owner of both the unfinished 66-foot 45 Park Place tower at the intersection of New York’s Tribeca and Financial District neighborhoods and the vacant lot next to it, the company run by Sharif El-Gamal started running into financial troubles after lower numbers of apartments in the property have gone under contract than expected and overseas lenders started to cut off funding.

The initial plan to put a 71-foot-tall, 16,000-square-foot Islamic cultural and prayer center adjacent to the main tower also caught national attention around 2010 due to its proximity to the Ground Zero site of the former World Trade Center and was eventually scrapped amid controversy.

Related: A luxurious Canadian hotel brand is coming to snowy Japan

Amid both rising debts and multiple creditor efforts to foreclose, construction on the tower stopped in 2019, and the tower stood in a half-finished shell over the last seven years in one of the most well-recognized examples of a stalled project in New York and the U.S.

According to the bankruptcy filing, Park Place Partners Development reported $15 million in assets and $13.8 million in liabilities. Kevin Nash of the Goldberg Weprin Finkel Goldstein law firm is representing El-Gamal and Park Place Partners in the bankruptcy case.

Early renderings presented 45 Park Place as an ambitious luxury condo tower project.45 Park Place

How 45 Park Place became the most famous unfinished condo tower in New York

Park Place Partners Development could not be reached for comment on the filing; as a result, little other information on why the company filed for bankruptcy now, after years of setbacks, is currently publicly available.

Amid multiple foreclosure efforts, El-Gamal has previously threatened to tear down the top of the 667-foot tower that has already been built in order to make the project less valuable and “get back” at creditors that he felt were purposefully preventing him from getting ahead.

More Travel News:

Airline to launch unusual new flight to Cayman Islands from the U.S.

There is a very cool Irish version of swimming pigs in The Bahamas

Unexpected country is most luxurious travel destination for 2026

Low-cost airline launches easier way to get to Sri Lanka

In response, the lending group that included creditors like Malaysia’s Malayan Banking Berhad, Kuwaiti Warba Bank and Chicago-based MSD Partners accused El-Gamal of creating a “bad faith scheme” to “gain leverage in negotiations.”

“To the extent that Mr. El-Gamal and his companies seek to pursue this scheme, they will be held fully liable for any and all damages,” the group wrote in a letter through the attorneys representing them as reported by The Real Deal in 2020.

Related: Another national park closes hotels, campgrounds, overnight parking

For years, figuring out whether to redeem loyalty points or miles for a trip has been tedious, frustrating and a real time zapper.

Do you use miles? Pay cash? Transfer credit card points? And are you actually getting a good deal when you redeem those hard-earned rewards?

That’s why I became a fan of Points Path when it first launched. The free browser extension made it much easier to compare cash prices with airline miles right inside Google Flights.

But Points Path has gotten even better. You can now use it to compare hotel prices, evaluate certain Chase Ultimate Rewards redemptions through Chase Travel and even use Points Path on an iPhone or iPad.

To me, Points Path is a travel game-changer.

What Is Points Path?

Points Path is a travel tool designed to help answer one of the hardest questions when using travel rewards: Should I pay with cash or points?

It started as a browser extension that works with Google Flights. Once installed, Points Path adds award prices alongside the cash prices you already see when searching for airfare.

Instead of searching Google Flights for the cash price and then opening multiple airline websites to see what those same flights cost in miles, you can see much of that information in one place.

Points Path has since expanded beyond airfare. You can now use it with:

Google Flights to compare cash and airline award prices.

Chase Travel to evaluate certain Ultimate Rewards redemptions.

Hotel booking sites to compare rates for the same property.

Safari on an iPhone or iPad through the Points Path app.

The basic Points Path service is free, although the company also offers paid features.

How Does Points Path Work With Google Flights?

This is still one of my favorite Points Path features.

Once you install the extension, search Google Flights as you normally would. Points Path adds the number of miles you would need to book eligible flights alongside Google’s cash prices.

Points Path supports a growing list of airline loyalty programs, including major programs from American, Alaska, Delta, JetBlue, United and Air Canada, along with a number of international airlines.

But Points Path doesn’t just show you the award price. It also helps you determine whether using miles or paying cash offers the better value.

Points Path does that by comparing the cash price with its estimated value of the miles required for the flight. The result is a quick recommendation to “Use points,” “Use cash” or “Use either.”

Points Path bases those recommendations on its own valuations for each airline’s points or miles. You can see Points Path’s current points and miles valuations on its website.

That doesn’t mean you always have to follow its recommendation. Maybe you’re sitting on a mountain of miles and would rather save your cash. Or perhaps you’re saving those miles for a bigger trip.

But Points Path gives you the information you need to make that decision without doing nearly as much homework yourself.

Points Path Example: Atlanta to Paris

Here’s an example of why I like this tool so much.

I searched Google Flights for a nonstop, round-trip flight from Atlanta to Paris in January 2028. The lowest nonstop cash fare was $2,063 round trip.

But with Points Path installed, I could also see the award prices — and, even better, Points Path gave me a quick recommendation on whether to “Use points,” “Use cash” or “Use either.”

For one Delta flight, for example, I could pay $2,063 or redeem 175,000 SkyMiles plus $136 in taxes and fees. Points Path labeled that redemption “Use either,” meaning the value of using miles was close enough to paying cash that either option could make sense.

Other flights in my search were labeled “Use cash,” making it easy to spot the redemptions that wouldn’t give me particularly good value.

I especially like this feature because you don’t have to do the points-versus-cash math yourself. At a glance, you can see the cash fare, available award price and Points Path’s recommendation.

Maybe you’d rather keep more than $2,000 in your bank account and use the miles, even if the redemption isn’t an outstanding deal. Or maybe you’re saving those miles for a trip where they’ll stretch further.

Either way, Points Path gives you a helpful starting point.

From there, you can select the flight and follow Points Path to the airline’s website to review the available award options and complete your booking.

Take your hands off the keyboard!

You may see several screen changes while Points Path does its work and redirects you to the airline.

Just remember that airfare and award prices change constantly, so your results will depend on when you search.

Have Fun With It!

One of my favorite things about Points Path is simply playing around with possible trips.

You don’t even need to have enough miles in your account to search. That means you can use Points Path to get an idea of how many miles you might need for a future trip — whether you’re dreaming about the Caribbean, Europe or somewhere much farther away.

It can turn “I’d love to go there someday” into a much more concrete savings goal.

How Does Points Path Work With Chase Travel?

Points Path can also help eligible Chase cardholders evaluate redemptions inside Chase TravelTM.

Chase’s Points Boost program can give eligible cardholders increased value for their Ultimate Rewards points on select flights and hotels. But the value can vary from one booking to another.

That’s where Points Path comes in.

When you search Chase Travel using a supported web browser, Points Path can calculate the value you’re getting per Ultimate Rewards point on eligible Points Boost redemptions. This can help you decide whether using points through Chase Travel offers a good value.

That’s particularly helpful because a redemption labeled “Points Boost” isn’t automatically the best possible use of your points. You may get more value by transferring your points to an eligible airline or hotel partner — or you may decide the convenience of booking through Chase Travel is worth it.

The Points Path Chase Travel feature works with both flights and hotels.

You’ll need to enable the Chase Travel feature through your Points Path account. Go to “Additional Permissions” to enable Chase Travel browser permissions. You’ll need to activate it on each device where you want to use it.

One important note: Points Path information appears when you access Chase Travel through a supported web browser. You won’t see the Points Path overlay inside the Chase mobile app.

Can Points Path Help You Compare Hotel Prices?

Yes. Points Path isn’t just about airline miles anymore.

The tool has expanded into hotel price comparisons, which could save you from opening a bunch of tabs before booking a room.

When you’re viewing a hotel on a supported online travel agency, Points Path can look for rates for the same property from other booking sources. You can also use its Compare Price feature to check prices elsewhere, including rates available directly from the hotel.

That’s useful because the price for the same room can vary depending on where you book it.

Of course, the lowest advertised price isn’t the only thing to consider. Before booking, compare:

Taxes and resort fees.

Cancellation and refund policies.

Room types and included amenities.

Hotel loyalty points and elite-status benefits.

Credit card travel protections.

Booking through a third-party site may mean you won’t earn hotel points or receive your usual elite benefits. Still, I love the idea of having a quick price comparison available while I’m already shopping.

You’ll need to grant additional browser permissions to enable the hotel feature. As of this writing, Points Path says hotel comparisons work best on a desktop or laptop and aren’t fully supported on iOS.

Can You Use Points Path on Your Phone?

Yes! This is another big change since Points Path first launched.

Points Path now offers an app for iPhone and iPad. But there’s an important distinction: It isn’t a standalone travel-search app. The app allows you to install and use the Points Path extension in Safari on an iOS device.

After downloading Points Path from Apple’s App Store, follow its instructions to enable the extension in Safari. Then you can see Points Path results while searching Google Flights in your mobile browser.

A couple of limitations to keep in mind: You will need iOS or iPadOS 18 or later, and not every desktop feature is fully supported on mobile. For example, Points Path says its hotel comparison feature is currently optimized for desktop and laptop browsers.

Android users are still out of luck when it comes to a Points Path mobile extension. Points Path says browser extensions aren’t currently available on Android.

What Browsers Support Points Path?

On a computer, Points Path is available for Safari and through the Chrome Web Store for Google Chrome and many Chromium-based browsers, including Microsoft Edge, Opera, Brave and Arc.

On an iPhone or iPad, you can download the Points Path app and enable the extension in Safari.

That makes Points Path much more accessible than it was when I first started using it.

Is Points Path Free?

Yes. You can install and use the basic version of Points Path for free.

Points Path also offers paid options with additional features for more serious points-and-miles users. But you don’t need to pay just to see basic Points Path results alongside your Google Flights searches.

I’d start with the free version and see how often you use it before deciding whether any paid features are worth the money for you.

Final Thoughts

Points Path started out solving one very specific travel headache for me: Should I use miles or just pay for the flight?

Now it can help answer quite a few more questions.

Is that Chase Points Boost redemption really a good deal? Is the hotel cheaper somewhere else? How many miles should I be saving for that trip I’ve been dreaming about?

And you can answer those questions without jumping from website to website or doing all the math yourself.

You still need to pay attention to the details before booking — especially taxes and fees, cancellation policies and any hotel loyalty benefits you could give up by booking through a third party.

But anything that makes travel rewards easier to understand and saves me time is a win in my book.

Points Path does both.
The post Points Path Review: This Free Tool Makes Using Travel Points Much Easier appeared first on Clark Howard.