Editorial Team
5 min read
Framework has previewed a Desktop built around AMD’s Ryzen AI Max+ PRO 495 and 192GB of unified LPDDR5X memory. The extra capacity could hold a large quantized model on one compact machine, but the processor change is modest, a rival system is planned, and Framework has not announced a price, ship date, or independent inference results.
Framework, the modular-PC maker whose Desktop expanded its lineup beyond laptops last year, has previewed a higher-memory version of that compact machine. At AMD’s Advancing AI event, the company said the forthcoming system pairs AMD’s Ryzen AI Max+ PRO 495 processor with 192GB of unified LPDDR5X memory—the first such combination it knows of. A separate account identifies last year’s Desktop as Framework’s first product other than a laptop.
The announcement is consequential less because of a new processor tier than because it lifts the amount of memory shared by the CPU and integrated graphics. That potentially lets an owner load a larger quantized language model locally. It does not yet establish what the machine will cost, when it will ship, or how quickly it will generate output.
Framework’s company-reported Desktop memory configurations, including the 192GB version listed as coming soon. Source: Framework.
The Framework Desktop is a small, configurable system built around a Mini-ITX mainboard, FlexATX power supply and 120mm CPU fan, according to Framework’s product page. Its existing AMD-based options reach 128GB of LPDDR5X-8000 memory; the coming configuration is listed with 192GB, 273GB/s of memory bandwidth, a 16-core/32-thread Zen 5 CPU, a 40-CU Radeon 8065S integrated GPU and up to 131 TOPS of AI compute.
The generational comparison matters. The PRO 495 retains the previous 395-class chip’s 16 Zen 5 cores and 40 RDNA 3.5 compute units. An independent technical account describes 100MHz increases to CPU and GPU clocks, while the new Gorgon Halo platform raises the supported LPDDR5X maximum from 128GB to 192GB.
That is 64GB, or 50%, more capacity. It is a meaningful difference for memory-bound local inference, but not evidence that all workloads will run faster. The published specifications give bandwidth and peak-compute figures; they do not give a token-per-second result for the model Framework is using in its pitch.
Framework says the configuration can run DeepSeek-V4-Flash at Q8 on a single system while leaving room for a longer context window. DeepSeek-V4-Flash is the language model used in the company’s example, not a benchmark of every model or workload.
One account of the configuration puts the model at 284 billion total parameters, with 13 billion active parameters, and says a Q8 GGUF build occupies about 162GB. Those reported figures explain the memory arithmetic: compared with the model file alone, 192GB leaves roughly 30GB before accounting for operating-system use, inference runtime and context. They do not validate a particular speed.
The company also plans a prebuilt Linux option. Its desktop remains aimed at users willing to configure or build a system, but the prebuilt offering could reduce one deployment step for local-model users. Framework describes the Linux distribution as pre-loaded; it has not set out the final software configuration in the preview.
The price question is not incidental. Framework says rising LPDDR5X prices mean the 192GB model will be a “pretty substantial” jump over its current 128GB Desktop pricing, while advising budget-conscious buyers to consider its 32GB, 64GB and 128GB systems. It has announced neither a price nor availability.
The available comparison base is imperfect but useful. A published price report cited $3,449 for the existing 128GB Ryzen AI Max+ 395 Framework Desktop in the U.S. That is a reference point rather than a quoted future price: neither Framework nor the report supplies the 192GB configuration’s final retail figure.
Nor is the 192GB pairing likely to be unique for long. GMKtec has confirmed an EVO-X3 with the same processor and memory capacity for later this year, according to a report. Framework’s claim is expressly limited to being first “as far as we know,” not to exclusive access to the platform.
Framework has also made the Desktop’s PCIe x4 slot open-ended, meaning physically longer cards can fit while the connection still carries four PCIe lanes. In its proof of concept, the company used 50GbE network cards and RDMA over Converged Ethernet to connect two 192GB systems, then used tensor parallelism to work across 384GB of memory.
That is a two-machine cluster, not a single 384GB computer. The distinction brings networking and software requirements that a one-box setup does not have. The demonstration nevertheless shows the adjacent use case Framework is courting: buyers for whom the single-machine memory ceiling is only the first constraint.
Framework has established the hardware ceiling and a plausible example of why it matters. The next evidence must be commercial and operational: a final price and ship date, the usable memory available after the system is running, and reproducible inference speed for the quoted Q8 model at stated context lengths.
Until those are available, the 192GB Desktop is best understood as a higher-capacity local-inference option—not proof that a compact PC has displaced the alternatives for large-model deployment.
Founded in 2025 by a team with autonomous-driving data experience, GenRobot AI says it raised $100 million in a Series A+. Its robot-data business reports million-hour quarterly deliveries, while funded rivals and undisclosed commercial metrics complicate the growth story.
AI safety advocates gained experienced lawmakers and battleground seats in the midterms, but rival technology-backed groups also elected allies, leaving federal legislation far from assured.
A federal judge found Anthropic’s Claude Max usage advertising plausibly misleading because subscribers could read the multipliers as monthly promises. Contract and misrepresentation claims survive, but several other claims were dismissed with leave to amend.
Thailand’s proposed data-centre framework would scale requirements by power demand, screen sites for electricity and water access, and assess domestic benefits. It also covers existing facilities, while conflicting application and approval schedules remain unresolved.
PJM’s independent market monitor attributes about $11 per megawatt-hour of first-half 2026 wholesale costs to data-center demand in capacity auctions. Energy costs drove more of the overall increase, while construction barriers and an October regulatory rejection complicate the response.
DataVita now says its proposed Lanarkshire AI data centres will be fully grid served, with direct renewable connections only a future possibility. The change in public presentation raises questions about promised power costs and community benefits, although the DV4 planning application already assumed grid supply.
AMD chief Lisa Su calls the chipmaker a data-center-first company as server and AI sales surge. Gaming revenue fell 40%, but PC-chip sales grew, and her remarks leave the supply outlook for expensive Radeon cards unresolved.
A reported U.S. intelligence assessment links Megaspeed executive Alice Huang to China’s Ministry of State Security, raising questions about access to Nvidia’s leadership and chip-export policy. Whether the alleged operation succeeded remains unclear.
Ocient’s $100 million Series C backs tailored AI analytics on an established database platform, with national-security investors and deployment economics shaping its expansion.
Anthropic plans IPO marketing next week and a listing in the week of November 16, according to people familiar with the plans. Buyers are discussing valuations of $1.2 trillion–$1.3 trillion, with some weighing $1.1 trillion, as its filing reveals faster-growing compute costs and extensive infrastructure commitments.
India's IT lobby chief warns that AI could make technology jobs harder for graduates to secure. Hiring is below earlier peaks, but some firms favour AI-equipped beginners over costly experienced staff. The unresolved question is how junior workers will build expertise.
Internet scans and leaked databases expose the spread of ARTEX, a security-testing framework used in cyberattacks. Researchers documented largely autonomous compromises, but the records also show failed cashouts and uncertain counts of independent deployments.
Meta's strategy chief and former CFO is joining Fireworks AI ahead of a reported fundraising push targeting at least a $30 billion valuation. The inference company announced a $17.5 billion valuation in July, making the hire part of a much larger expansion than its older funding figures suggest.
Sand.ai says its $200 million Series A+ round lifts its valuation above $1 billion. Co-lead investor Wangsu plans inference and deployment cooperation, while its filing shows substantial losses behind the startup’s video AI ambitions.
An advocacy group estimates Interior Secretary Doug Burgum’s AI holdings at $18.1 million–$64.4 million as his department advances data centers on public land. His ethics agreement allows retained investments with recusals, while pending permits will determine both the application of those safeguards and the public’s return. Interior denies his decisions are driven by personal investments.
Microsoft is evaluating fewer neocloud partners while expanding its own infrastructure. Delivery-dependent contracts underpin supplier financing, while IREN reports rising revenue alongside a $6 billion funding target for its broader expansion.
Physical Superintelligence says its AI system found a proof of a conjecture limiting quantum query advantages on random inputs. Public proof files enable scrutiny, but do not establish the private system’s cost or repeatability.
OpenAI is expanding its relationship with Hacktron, whose researchers previously reported an attack chain into employee accounts, to test third-party connections with greater access to cybersecurity models.
Google has acquired Crusoe’s southeast Wyoming data center site, where it plans a long-term $17 billion investment. Crusoe describes completed pre-development work; independent reporting points to customer setbacks. Power delivery and carbon capture remain key tests.
The team behind document-parsing service SoMark has released OXR-1.0 code and weights. A benchmark retest supports a strong result, but speed and accuracy were measured separately, and users above funding or revenue thresholds need a commercial license.