The first surface humankind carved memory into. Now it carves memory for AI.
“Carved like stone: light, fast, and the same in every language.”
Can be asked again when one pass is not enough, and stores images. 95.7% on LongMemEval-S, 67.5% on BEAM 1M. Same price as Tablet 1.
See Tablet 2The first publicly available WOS model: the memory engine you can call today. On LongMemEval-S it scores a mean of 85.2% across five runs, returning about 1,200 tokens of context per query in roughly 320 ms. No LLM in the path, and the same behaviour in every language.
Full benchmark report