MemryX says MX3+ will deliver up to 10 times MX3 compute
Co-founded by Michigan professor Wei Lu, MemryX says MX3+ targets transformers and small language models, with samples targeted for late 2027.
By RuntimeWire Staff · Published
Primary source: PR Newswire Business Technology
Why it matters
Edge customers are asking cameras, robots and factory systems to run transformers and small language models on tight power budgets. MemryX's answer is a 16-chip design its compiler treats as one device, which, if the late 2027 samples match the targets, would let existing MX3 customers scale up without redesigning around a new architecture.
Co-founded by Michigan professor Wei Lu, MemryX says MX3+ targets transformers and small language models, with samples targeted for late 2027.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.