Framework has revealed its next generation mini desktop built around AMD top tier Gorgon Halo silicon. The compact machine packs a massive 192GB of unified memory and custom Noctua cooling specifically tuned for running heavy artificial intelligence models locally. But with steep component prices pushing top tier configurations toward $6800, getting your hands on this hardware will cost a fortune.
Under the hood sits the 16 core AMD Ryzen AI Max+ Pro 495 processor featuring integrated Radeon 8065S graphics with 40 compute units. Framework paired the chip with 192GB of soldered LPDDR5X 8533 unified memory pushing 273 GB/s of bandwidth. The company claims this hardware delivers 131 TOPS of total computing power. That is enough throughput to run massive local language models like DeepSeek V4 Flash without touching a cloud server. Users can even link 2 separate units together into a single cluster to reach 384GB of shared memory.
The external layout uses the exact same mini ITX footprint as the previous generation to maintain backwards compatibility. Cooling duties fall on a massive Noctua NF A12x25 fan to keep temperatures stable under continuous processing loads. You get a single open PCIe x4 slot for networking cards or low profile expansion boards alongside a standard Flex ATX power supply. Buyers can order the system preinstalled with either Windows 11 or Fedora Linux.
Skyrocketing component prices mean this modular workstation carries a steep barrier to entry. Addressing customer reactions over the expensive price tag on its official X account, Framework kept the explanation blunt.
Memory costs are a mess.
While previous generation setups with 128GB of memory started around $3450, early retailer listings show the top tier 192GB models pushing all the way to $6800. For developers who need massive local memory pools without buying a bulky server rack, this desktop remains one of the few compact options on the market.
