📰 News

  • July 2026: We introduce Macaron-V1, our open agent-model family for experiential intelligence, available in 748B Venti and 35B Tall variants and built on Mixture-of-LoRA and model-harness co-design. Learn more: Introducing Macaron-V1

  • May 2026: Our paper δ-mem: Efficient Online Memory for Large Language Models is out, a lightweight memory mechanism that augments a frozen backbone with a compact online associative state, improving memory-heavy benchmarks while preserving general capabilities.

  • January 2026: We introduce MinT, our infrastructure for training and serving millions of LLMs that enables experiential intelligence at scale. Learn more: MinT: RL Infrastructure for Experiential Intelligence

  • December 2025: We launch Mind Lab 🎉, a research lab focused on Experiential Intelligence, building AI systems that improve from real-world experience rather than just optimizing fixed benchmarks, moving from static “brains” to adaptive “minds”. Learn more: Introducing Mind Lab: Building AI that Learns from Real Experience