Exabase M-1: Cheaper AI Model Tops Memory Benchmark

Exabase's new memory engine, M-1, has achieved the highest reported score on LongMemEval, a leading benchmark for conversational AI memory. M-1 scored 96.4% accuracy, outperforming all other systems, including Mem0, Honcho, HydraDB, and Supermemory. Here's the thing: M-1 accomplished this using Gemini 3 Flash. This model is four to six times cheaper and faster than Gemini 3 Pro, which other top systems used. What's interesting is that M-1's memory architecture allows it to achieve better results with a less expensive model. The founder of Exabase stated that this makes the difference between a benchmark result and a production system. The bottom line is that this development could mean more efficient and cost-effective AI agents for various applications.