FrontierMemory v1.1
Agents · 2026-10-29
Letta's benchmark of memory generation. A model reads 750 past agent trajectories and writes a memory, which a fixed task agent then uses on 15 coding and knowledge-work tasks that need that knowledge. Each row is the mean over three independent runs of the higher full-pass rate of Opus 5.5 and GPT-6 Sol at max reasoning, with memory generated in Letta Dream and tasks run in Letta Code. Corpus v1.1 replaced the 665-trajectory corpus of v1.0.
Benchmark source: Frontiermemory