Xiaomi API
Xiaomi's MiMo line is served through the Xiaomi MiMo API Open Platform, which accepts OpenAI- and Anthropic-style requests. Felo currently lists one route, MiMo-V2.6-Flash, an MIT-licensed multimodal model built for high-frequency office, coding and agent work. Felo calls it with a Felo API key, so no separate Xiaomi platform account is required.
- Models
- 1
- chat ×1
- Input per 1M
- $0.14
- Output per 1M
- $0.28
- Largest context
- 1.05M
What Xiaomi publishes under MiMo
The MiMo platform's current generation is V2.6: MiMo-V2.6-Pro for maximum reasoning, MiMo-V2.6-Flash as the balanced, low-cost tier, and MiMo-V2.6-Pro-UltraSpeed for up to 20x output speed. Alongside the chat models, Xiaomi publishes speech models for text-to-speech and speech recognition and supports image, audio and video understanding.
Xiaomi publishes the weights for the family, and the Felo catalog lists MiMo-V2.6-Flash under the MIT license; developer tools around it include MiMo Code, MiMo Claw and a desktop client.
Access and billing
Developers sign in to the MiMo API Open Platform with a Xiaomi account, then choose between pay-as-you-go billing and a Token Plan; team Token Plans add seat, usage and billing management. Credentials differ by usage mode, and the console lists model IDs and current prices.
Batch inference is available for large jobs that do not need real-time responses, and the platform documents text generation, tool calling, web search, deep thinking and structured output as standard capabilities.
Compatibility and migration
The platform documents both OpenAI API and Anthropic API compatibility, so existing SDK clients can be pointed at MiMo with changed credentials and a different base address. Model IDs are versioned: MiMo-V2.5-Pro and MiMo-V2.5 were retired on October 21, 2026, and integrations should use the V2.6 series.
Where MiMo fits
MiMo-V2.6-Flash is aimed at high-frequency office calls, large-scale tasks, coding and agent work, and visual understanding, with text, image, video and audio input, a 1,050,000-token context and up to 131,072 output tokens. It is the cost-first option: the catalog positions it below the Pro tier on price while keeping near-Pro results on coding-agent benchmarks. Two limits: the MiMo platform currently offers personal-account login only, and Felo carries a single MiMo route rather than the full V2.6 family.
Specifications
- Input per 1M
- $0.14
- Output per 1M
- $0.28
- Context
- 1.05M
- Max output
- 131.072K
- Protocols
- messages
Similar models from other providers
Frequently asked questions
- Which Xiaomi model is available in Felo?
- One route: MiMo-V2.6-Flash, a chat model that accepts text, image, video and audio input, returns text, and offers a 1,050,000-token context window with up to 131,072 output tokens.
- Is the MiMo API OpenAI-compatible?
- Yes, and it also accepts Anthropic API format. Xiaomi documents both so existing SDK clients can switch with changed credentials and a different base address.
- How is Xiaomi MiMo billed?
- The MiMo platform offers pay-as-you-go token billing and Token Plan subscriptions for teams. The table on this page shows the published USD token rates for the route Felo carries.
- What happened to MiMo-V2.5?
- Xiaomi retired MiMo-V2.5-Pro and MiMo-V2.5 on October 21, 2026, and their model IDs no longer work. New integrations should target the V2.6 series.
- Can MiMo weights be self-hosted?
- MiMo-V2.6-Flash is listed under the MIT license in the Felo catalog, so its weights can be self-hosted; check each model's release notes for the license of other variants.
- Does MiMo accept audio and video input?
- The Flash route accepts text, image, video and audio input and returns text. Xiaomi's docs also cover image, audio and video understanding and speech models for TTS and ASR.