Xiaomi stuns with new MiMo-V2-Pro LLM nearing GPT-5.2, Opus 4.6 performance at a fraction of the cost (10 minute read)

Xiaomi's MiMo-V2-Pro is a 1-trillion-parameter foundation model with performance approaching that of models from OpenAI and Anthropic but at a fraction of the cost. The model uses a sparse architecture that only activates 42B parameters during any single forward pass. It has a Multi-Token Prediction layer that allows it to anticipate and generate multiple tokens simultaneously, drastically reducing the latency required for 'thinking'. The model is currently only available via Xiaomi's first-party API. Xiaomi plans to release an open source variant of the model.