Back to events
Open-source releaseModerate global significanceConfirmed confidence

Xiaomi releases MiMo V2.5 multimodal agent model

Xiaomi releases MiMo V2.5 for native text, image, video and audio understanding; model weights follow under the MIT license on June 29.

Event details

Xiaomi introduced MiMo V2.5 on April 23, 2026, as an autoregressive Transformer for multimodal agent use. Its parameter count was not disclosed. The model weights were released under the MIT license on June 29.

Why it matters

MiMo V2.5 expands Xiaomi's agent-model line into native understanding across text, images, video and audio, with permissively licensed weights.

53/100Global significance score. Regional effects are recorded only when the evidence supports a meaningful difference.

Access notes

Introduced April 23, 2026; model weights released under the MIT license on June 29.