FrontierMemory v1.1 Harnesses
Agents · 2026-10-29
Letta's comparison of memory systems on the same 15 tasks: Letta Dream, Hermes Agent, Claude Managed Agents, Prime Agent and OpenAI Agents SDK, each with its native task harness and a matched model pair. It ranks systems rather than models, so it is a diagnostic.
Benchmark source: Frontiermemory
Top models (higher is better)
| Model | Score | Source |
|---|---|---|
| Claude Opus 5.5 | 73.3 | FrontierMemory v1.1 harnesses: Letta Dream · Letta |
| GPT-6.1 Sol | 26.7 | FrontierMemory v1.1 harnesses: OpenAI Agents SDK · Letta |