MiMo V2.6 Flash lands on Inworld

MiMo V2.6 Flash becomes a compelling low-cost option for multimodal agents and long-context workloads, though developers should validate reliability before treating it as a drop-in replacement for leading closed models.

• –OpenAI-compatible APIs reduce migration friction for existing applications.
• –Cached input at $0.003 per million tokens makes repeated long-context workflows especially economical.
• –Vision and function calling expand its usefulness beyond text-only reasoning.
• –Xiaomi’s broader V2.6 release includes open weights, RL resources, and 3D game-development capabilities. [Xiaomi](https://mimo.mi.com/docs/en-US/news/latest/v2-6)
• –Inworld lists a 1M-token context window, matching Xiaomi’s advertised maximum.