After the Export Ban: How Meta's 816B-Parameter Muse Spark Open Redraws the Open-Weight AI Race
Friday, October 16, 2026 · 32 items · 6 min read · Updated 1:03 AM
By the Numbers
120 billion
parameters in Nemotron 3.5 Super VL
AI
$115.7 million
Q3 revenue
Critical Minerals
23.3%
year-over-year growth
Critical Minerals
$334.7 million
nine-month revenue
Critical Minerals
4,000
personnel assigned to vulnerability remediation
AI
1,000
outside contractors hired
AI
The Day's Thesis
▶
Signal of the Day: Meta's Muse Spark Open, an 816-billion-parameter open-weight model processing text, images, video, and audio with up to one million tokens of context, launches optimized for Apple's M5 Ultra chip — matching top Chinese open-weight models on benchmarks.
The 30-Second Read:
Meta's Muse Spark Open (816B parameters) benchmarks level with Qwen, GLM, and Kimi, the leading Chinese open-weight models
CoreWeave credit default swaps — insurance-like contracts that pay out on debt default — hit distress levels, signaling market deterioration in AI infrastructure debt
Google's Gemini 4 Argon scores 87% on its ERQA robotics benchmark, positioning it as a single-model stack for physical AI
Freegold drills 1.24 g/t Au over 213 meters at Golden Summit, anchoring a 2027 pre-feasibility study timeline
Today's dominant tension is a bifurcation in AI infrastructure confidence: frontier model capability is expanding rapidly at the open-weight layer, while the debt markets financing the compute layer are flashing distress. These two facts are not unrelated — they share the same capital cycle.
AI & Research Frontier
Meta's Muse Spark Open puts an 816-billion-parameter, million-token-context multimodal model into local deployment on consumer Apple silicon, directly challenging Chinese open-weight leadership.
The model processes text, image, video, and audio natively and is optimized to run on Apple's M5 Ultra — a pairing that eliminates cloud dependency for a class of workloads previously requiring data center infrastructure.
On standard open-weight benchmarks, Muse Spark Open performs roughly on par with Qwen, GLM, and Kimi, the three dominant Chinese open-weight models. The competitive parity is strategically significant: it resets the baseline for what Western open-weight AI can deliver without API access or cloud spend.
MIT's Center for Transportation and Logistics finds sustainability commitments are strong, but governance gaps and external disruptions prevent execution.
Separately, a new computational model from academic researchers decoded intercellular signaling — the chemical language cells use to coordinate tissue assembly. The approach could compress stem cell engineering timelines and improve disease modeling fidelity, representing an early-stage application of AI to biology that sits upstream of any near-term commercial valuation.
The plain-language version: Meta just released a free, locally-runnable AI model powerful enough to compete with China's best open-source systems.
Technology & Infrastructure
OpenAI's ChatGPT Chrome extension now reads users' open browser tabs on demand, expanding active-session context without requiring a dedicated browser product.
The capability — demonstrated via a public video showing cross-tab sofa comparison — represents a shift from passive to active browser integration. It arrives after OpenAI discontinued the ChatGPT Atlas browser, suggesting a pivot toward extension-layer access over standalone browser development. The specific permissions and data-handling architecture have not been disclosed, which will matter for enterprise adoption decisions.
CoreWeave credit default swaps reaching distress levels is the more consequential infrastructure signal of the day. CDS spreads at distress thresholds indicate that bond market participants are pricing a materially elevated probability of debt restructuring or default — this is not equity volatility but credit market stress, which has different and typically more durable consequences for capex cycles. CoreWeave carries significant debt load from its GPU cluster build-out, and distress-level CDS spreads will raise its cost of incremental borrowing precisely when the AI infrastructure cycle demands continued spending.
The plain-language version: The bond market is signaling it no longer trusts that CoreWeave's GPU debt will be repaid on schedule.
Markets & Capital Flows
CoreWeave's credit default swaps hitting distress levels — continuing the 15% share decline flagged in prior sessions — mark the sharpest single deterioration in the AI infrastructure debt complex this cycle.
The CDS move is a leading indicator: distress-level spreads precede formal restructuring conversations by weeks to months in typical credit cycles. For the broader AI infrastructure financing thesis, this matters because CoreWeave's balance sheet model — borrow at scale, deploy GPUs, collect hyperscaler contracts — has been the template for independent AI compute providers. Stress at CoreWeave raises the cost of that template across the sector.
Netflix's announcement that Chief Content Officer Bela Bajaria will be replaced by former Disney executive Peter Rice represents a significant leadership transition at a company spending over $17 billion annually on content — but carries no direct AI infrastructure or minerals implication today.
The Indian rupee's partial recovery from 96.79 to the 95-per-dollar range reflects Reserve Bank of India intervention, relevant to emerging market currency exposure but not directly connected to today's technology or minerals events.
The plain-language version: AI infrastructure debt is cracking, and the CoreWeave CDS move is the clearest single number proving it.
Critical Minerals & Supply Chain
Freegold's 213-meter intercept grading 1.24 g/t Au — including a 21-meter high-grade zone at 5.77 g/t — at Golden Summit defines the drill target for a 2027 pre-feasibility study, one of the more substantive junior gold results of the quarter.
The result is meaningful for resource definition rather than production — pre-feasibility studies — formal engineering assessments of whether a deposit is economically viable — require this kind of geometry to model mine design. The 2027 timeline means no production contribution before 2029 at earliest under an accelerated scenario.
MIT's Supply Chain Sustainability Report finds corporate commitment to sustainability at a recorded high, but identifies governance gaps and external disruptions as the primary execution barriers. The report does not quantify a dollar figure for the governance shortfall, limiting its near-term market utility — but it establishes a documented baseline against which 2027 supply chain disclosures will be measured.
OneGold launched a Mastercard that instantly liquidates gold and silver holdings at spot prices, backed by a vault of over 4 million ounces insured through Lloyd's of London. The product converts physical precious metal exposure into point-of-sale liquidity, a structural change in how retail gold is accessed.
The plain-language version: A new card lets you pay at any store directly from your gold holdings — no selling, no waiting, no fees on US conversions.
The Interconnect: Cross-Sector Causal Chains
→Meta's Muse Spark Open optimized for Apple M5 Ultra local deployment → removes cloud compute dependency for 816B-parameter inference → CoreWeave's hyperscaler contract revenue base faces incremental erosion as local silicon absorbs workloads previously requiring rented GPU clusters reported
→CoreWeave CDS spreads reaching distress levels → raises cost of debt capital for GPU cluster build-outs across independent AI compute providers → capex deployment timelines for non-hyperscaler AI infrastructure extend, reducing near-term demand growth for power infrastructure and associated copper and cooling equipment reported
→Google Gemini 4 Argon's 87% ERQA benchmark score for robotics control → positions a single foundation model as the software stack for physical AI systems → semiconductor demand shifts toward edge inference silicon — low-power, high-throughput chips designed for on-device processing — rather than centralized data center GPU clusters, with supply chain implications for packaging and memory reported
Watchlist
▸CoreWeave — CDS spread trajectory and any formal debt restructuring disclosure · Catalyst: Distress-level CDS confirmed by multiple market participants · When: Next 30–60 days
▸Meta — Muse Spark Open enterprise adoption rate and benchmark updates vs. Qwen 3 and GLM-5 · Catalyst: Open-weight release; Apple M5 Ultra optimization confirmed · When: Q4 2026 developer uptake data
▸Apple — M5 Ultra silicon demand signal from Muse Spark Open local deployment · Catalyst: Meta partnership for on-device 816B-parameter inference · When: Touchscreen MacBook launch, days away per reporting
▸Freegold — Pre-feasibility study scoping parameters following 213m/1.24 g/t drill result · Catalyst: Golden Summit intercept defines 2027 PFS starting geometry · When: 2027 PFS initiation, interim resource update likely H1 2027
▸OneGold / Bullion International Group — Transaction volume through new Mastercard product and fee-waiver uptake · Catalyst: Card launch with 4M+ oz vault base and $1.5B transaction history · When: December 31, 2026 silver fee-waiver expiry is the first hard date
▸MIT Supply Chain Sustainability Report — Corporate governance response and whether 2027 disclosures close the execution gap identified · Catalyst: Documented baseline established October 2026 · When: Annual report cycle, Q1 2027
Dr. Mark Thornton, senior fellow at the Mises Institute, shares his outlook for gold and silver, laying out the factors that have been weighing on prices and commenting on what may kick off their next leg higher — in his view, the return of speculators could provide a boost.
"Speculators we found out last year are a main engine to the prices of gold and silver," he said.
Don’t forget to follow us @INN_Resource for real-time updates!
Securities Disclosure: I, Charlotte McLeod, hold no direct investment interest in any company mentioned in this article.
A new computational approach reveals how cells communicate to assemble into tissues. This could accelerate stem cell engineering, improve disease models, and inform therapies.
OpenAI's ChatGPT Atlas browser is gone, but now ChatGPT's Google Chrome extension is getting some upgrades that allow it to use the browser tabs you already have open. With the extension installed, OpenAI says ChatGPT can now view your browser tabs when you tag Chrome in ChatGPT's web or desktop app, posting a demo video showing someone asking ChatGPT to compare the three tabs they have open with sofas they're considering.
A few big upgrades for Chrome users today! We're bringing the power of computer use to anyone who uses Chrome.
First up: ChatGPT can now use your tabs in Chrome to make answers more useful and get things done for you.
If …
Read the full story at The Verge.
Singapore continues to grapple with severe haze that is disrupting daily life and businesses, as authorities roll out protective measures — and issue warnings.
Meta's Muse Spark Open is a new open-weight model with 816 billion parameters that processes text, images, video, and audio with up to one million tokens of context. It performs roughly on par with the most capable open-weight Chinese AI models on benchmarks and is optimized to run locally on Apple's M5 Ultra.
The article Muse Spark Open is Meta's answer to Qwen, GLM, and Kimi in the open-weight race appeared first on The Decoder.
This week on “Uncanny Valley,” we dive into an unexpected Muse feature, shadowy spending from the largest pro-AI super PAC, and men’s tendency to lie about how much they know about AI.
Elon Musk gave a record $158 million to the GOP’s Senate Leadership Fund, boosting its war chest as Republicans pour money into Texas to defend Ken Paxton.
Google's Gemini 4 Argon controls robots, generates simulations from photos or sketches, and evaluates training videos. On Google's own ERQA benchmark, it scores 87 percent. But the published example prompts show that the demos still need a lot of human guidance.
The article Google wants Gemini 4 Argon to be the only AI model robots need appeared first on The Decoder.
President Donald Trump posted a series of photos on Truth Social showing off renovations to the White House - but one of them appears to be modified with AI. The picture, which compares an old White House column to a "restored" one, has several discrepancies and was flagged by Google's AI detector as "made or edited with OpenAI."
As pointed out by users on Reddit, the picture of the "restored" column has an uneven number of steps on both sides of the structure. The number of steps also doesn't match up with the ones in the "before" photo. Additionally, the before and after photos appear nearly identical in angle and detail; the reflection i …
Read the full story at The Verge.
Powerful AI models continue to cause chaos, but their targets aren’t widely seeking justice. Instead, they’re being encouraged to adopt those same tools.
Vaulted precious metals platform OneGold announced the launch of a new Mastercard (NYSE:MA) that instantly liquidates users' gold and silver holdings at spot prices to fund everyday retail purchases and ATM cash withdrawals.
The OneGold Card card was introduced alongside a full rebuild of the company's mobile application and website, allowing customers to spend their allocated metal balances anywhere Mastercard is accepted.
The system bypasses the traditional friction of liquidating physical bullion, eliminating the need to sell assets at coin shops, wait for checks to clear, or delay transactions while bank transfers settle.
When a customer swipes the OneGold Card or utilizes it via Apple Pay and Google Pay, the platform instantly converts the required fraction of their gold holdings into fiat currency at the current spot price.
The company stated it will not charge selling fees on US gold conversions used to fund purchases or withdrawals. A similar selling fee waiver applies to US silver holdings through December 31, 2026.
“With this card, gold is money again,” said Ken Lewis, CEO of Bullion International Group, OneGold's parent company. “For centuries, gold has been the ultimate store of value and inflation hedge, but it hasn’t been something you could spend at the register. The OneGold Card removes that final barrier, making precious metals as liquid as the cash in your wallet.”
Founded in 2018, OneGold attempts to combine the liquidity and low trading costs of an exchange-traded fund with the direct ownership of physical metal. Customers hold allocated positions in gold, silver, and platinum stored in their choice of vaults globally.
The platform currently manages more than 4 million ounces of metal, which is fully insured through Lloyd’s of London and verified by regular third-party audits. To date, OneGold has processed over US$1.5 billion in transactions.
Bullion International Group operates multiple entities across the precious metals supply chain, providing minting, vaulting, wholesale, and retail services through brands including APMEX, MTB, and GOLD AVENUE.
Don’t forget to follow us @INN_Resource for real-time news updates!
Securities Disclosure: I, Giann Liguid, hold no direct investment interest in any company mentioned in this article.
Editorial Disclosure: OneGold is a client of the Investing News Network. This article is not paid-for content.
Nvidia released Nemotron 3.5 Super VL, an open multimodal model with 120 billion parameters designed for enterprise AI agents, with dozens of organizations already fine-tuning it. The model's practical deployment requires purchasing Nvidia hardware, creating a direct hardware revenue lever tied to enterprise AI adoption.
Bluesky's Attie AI tool expanded beyond feed customization to enable users to create and share small applications like games, polls, and dashboards as standalone websites. This decentralizes app creation and distribution, potentially lowering barriers to entry for non-technical creators in the social platform ecosystem.
Tesla plans to reveal the next-generation Roadster after missing every targeted delivery date since Elon Musk's 2017 concept debut. Repeated delays signal either significant technical challenges or shifting internal priorities, both of which carry implications for Tesla's product roadmap credibility and investor confidence.
Capital, a London-listed mining company, reported Q3 revenue of $115.7 million, representing a 23.3% year-over-year increase and nine-month revenue of $334.7 million, up 32.2% year-to-date, with its drilling platform gaining operational momentum. Strong revenue growth indicates expanding exploration or extraction capacity in the critical minerals supply chain.
Bank of America deployed 4,000 people (including 1,000 outside contractors) to remediate software vulnerabilities uncovered by AI security models like Anthropic's Mythos. This large-scale defensive labor deployment demonstrates how AI-assisted vulnerability discovery is creating immediate, substantial remediation costs for major financial institutions.
Amazon devices chief Panos Panay outlined a vision where ambient AI devices anticipate user needs and operate in the background without requiring explicit app-based commands. This strategic shift toward proactive, context-aware AI represents a fundamental reorientation of how consumer devices mediate interactions with digital services.
KFC is switching its U.S. beverage supplier from PepsiCo to Coca-Cola for the first time in decades, marking a significant reversal in a long-standing partnership. This change affects a major customer relationship in the beverage industry and signals shifting commercial arrangements in the QSR supply chain.
Yukon Metals confirmed a new copper-gold-molybdenum porphyry-style mineral system at its Viper Prospect in Arizona, expanding the company's resource portfolio. The discovery of a polymetallic porphyry system adds both critical minerals (molybdenum) and base metals to Yukon's asset base, strengthening its position in the broader metals supply chain.
DailySand