Xiaomi is phasing out its MiMo-V2.5-Pro and MiMo-V2.5 models, officially retiring them on October 21st at 10:00 AM (UTC+8). This strategic move aligns with the continuous evolution and performance enhancements of the MiMo model series.
Transition to Advanced Models
The retired V2.5 models are being replaced by their successors, MiMo-V2.6-Pro and MiMo-V2.6-Flash. This transition signifies a significant upgrade, particularly for the Pro version. While MiMo-V2.5-Pro was limited to text input, the new MiMo-V2.6-Pro introduces comprehensive multimodal understanding capabilities, marking a substantial leap in its functionality.
Key Changes and Migration Details
As of October 14th, 2026, at 6:00 PM (UTC+8), requests originally directed to the V2.5 models have been automatically rerouted to their V2.6 counterparts. This ensures a seamless user experience during the upgrade process.
However, certain parameter handling rules have been updated:
- In the thinking mode, MiMo-V2.6-Flash and MiMo-V2.6-Pro do not support custom settings for
temperatureandtop_p. The system will default totemperature:1.0andtop_p:0.95.
For users migrating from MiMo-V2.5:
- If custom parameters were used with MiMo-V2.5, these settings will be inherited when routed to MiMo-V2.6-Flash.
- If
max_completion_tokenswas not specified in the request, the system will automatically adopt MiMo-V2.6-Flash’s default value.
Default Value Differences
A notable change in default values concerns max_completion_tokens:
- MiMo-V2.5 had a default of 32768 tokens.
- MiMo-V2.6-Flash now defaults to 131072 tokens.
Xiaomi’s MiMo-V2.6 Series: Performance and Pricing
Xiaomi officially launched and open-sourced the new Xiaomi MiMo-V2.6 series on September 22nd. This series comprises two native, fully multimodal models: Pro and Flash.
The MiMo-V2.6 series maintains the same API pricing structure as the V2.5 series, making advanced AI accessible:
- Flash version: Priced at 1 Yuan per million input tokens and 2 Yuan per million output tokens.
- Pro version: Priced at 3 Yuan per million input tokens and 6 Yuan per million output tokens.
Furthermore, users can benefit from a 99% cache discount. Xiaomi emphasizes that, at equivalent intelligence levels, these models are significantly more cost-effective than international alternatives, costing only 1/20th to 1/60th of their prices.
The MiMo-V2.6 series has already garnered attention for its performance, reportedly surpassing models like Kimi K3 and GLM-5.3 to become the highest-ranked open-source model on the AA index.
